How much do studios actually use AI? This page tracks which AI tools studios mention in their job postings, how often, and which roles already use them - measured from real hiring requirements, not opinions. A mention is not a usage requirement - it can be a "nice to have" or simply named in the about-us blurb. Still, the relative weight between tools is informative. Counts are all-time - across every posting tracked since launch, not just this week; the Dashboard shows the current week.
Every AI tool named in job ads, ranked by how many postings mention it (all-time). The number is the mention count; the bar is relative to the most-mentioned tool. Chips on the right are the roles that name it most - the % is the share of that tool's mentions coming from that role. A mention isn't proof of required use - it can be a "nice to have" or named in passing.
For every job ad that names an AI tool, we look at what craft skills the same ad also asks for. It shows the crafts AI is layered on top of - the human work studios still want next to the tool - not what it replaces. Each bar is the share of that tool's ads that also mention the craft.
▼AI adoption by discipline6 disciplines
Share of each discipline's job ads that name at least one AI tool - how widely AI shows up in its hiring. Beside each row: AI ads / total ads (the blue number is how many named a tool), then a chip with the discipline's most-named AI tool. Bars are scaled to the leader.
Tech Art
5.4%
Flux33 / 613
VFX
4.9%
Flux78 / 1605
Games
4.8%
Flux146 / 3051
Motion
4.8%
ChatGPT6 / 125
Animation
1.6%
ComfyUI1 / 61
Arch Viz
0.0%
0 / 1
▼AI adoption by countrytop 12
Share of each country's job ads that name at least one AI tool. Beside each row: AI ads / total ads (the blue number is how many named a tool), then a chip with that country's most-named tool. Countries with at least 10 postings only, so a tiny sample can't show a misleading rate. Bars are scaled to the leader.
▼AI tools on the movetrend over the last few weeks
AI tools on the move: the AI tools whose mentions in job ads changed the most over the last few weeks.
The blue number on the left is how many open postings name that tool right now. The line shows how that moved week by week. The arrow and % on the right are the overall trend - the slope of a line fitted through all the weeks (green = rising, red = falling).
Order: sorted by how much the mentions actually moved - postings gained or lost - biggest first.
▼Traction leaderboard · click to expand11 categories
How to read: within each category, tools are scored 0–100 by
capability coverage (40%), peer-relative adoption on HF + GitHub (25%),
mentions in tracked job postings (25%) and recency of public activity (10%).
A “-” in a column means we have no signal there
(e.g. closed-source tool has no GitHub) and its weight is redistributed
across the rest. Not a quality review - output style is not measured.
Each card shows tools in that category ranked by how much of the category's capability vocabulary they support. Yes = full credit, limited = half credit, no = 0. The orange bar marks the category leader. Use +N more to see the long tail.
Every tracked tool ranked by capability coverage in its primary category. Bars are not absolutely comparable across categories (different vocabularies) but show who leads their peer set. Use Select to keep only specific tools, Clear to reset. Top 20 visible by default - Show all reveals the long tail.
Weekly mentions of AI tools in postings, grouped by category.
Pick a category to see all of its tools as lines; refine with
Select. X-Ray overlays the prior window of
equal length as dashed lines.
Note:
data is still accruing - the chart will get richer as we
collect more weeks of postings.
Full catalogue of every AI tool we track, sorted by how often it is
mentioned in job postings. Badges: FREE - has a free
tier; SELF-HOST - can run locally; OSS
- open source; NO PUBLIC METRICS - a proprietary model
with no public HuggingFace / GitHub signals to measure, so its position
here reflects job-posting mentions only. A state-of-the-art closed model
(Claude, Nano Banana, Midjourney...) can sit low simply because we have
no popularity metric to rank it by - not because it is weak.
🔍
Fluximage-gen
Black Forest Labs
Flux is a family of text-to-image models from Black Forest Labs, founded by members of the original Stable Diffusion team, and it quickly became the open-weights favourite for image quality and prompt-following. It ships in tiers: a fast distilled Schnell release under a permissive licence, an open-weight dev model for non-commercial use, and a hosted Pro tier via API. Like Stable Diffusion it runs locally and plugs into ComfyUI, so studios can keep generation in-house and build ControlNet- and LoRA-style pipelines around it. In production it covers concept art, texture and matte ideation, and any work where local control and higher fidelity matter. Its stronger typography and coherence pushed many SD-era pipelines to migrate.
62 mentions
Claudetext
Anthropic
Claude is Anthropic's assistant family (Opus, Sonnet, Haiku), known for strong long-context reasoning, careful instruction-following and codebase work, with an emphasis on safety and reliability. In creative-tech it shows up for documentation, scripting and tools development, research synthesis and long-document analysis, and increasingly powers agentic coding via Claude Code. It is fully hosted with a free tier and no self-hosting; access is through chat apps and an API used to build internal tools. Tracked here because it appears in postings as a named AI assistant alongside ChatGPT and Gemini, reflecting AI literacy as a baseline skill - a general reasoning and writing aid for the people building and supporting productions rather than a direct art tool.
41 mentions
Aidercode
Open-source
Aider is an open-source command-line pair-programmer that edits a git repository using LLMs (Claude, GPT and others), making changes as proper commits with awareness of the whole repo. Because it is open-source and model-agnostic, developers use it for repo-aware refactors, scripted edits and automation without committing to a single vendor's editor. In creative-tech it suits pipeline TDs and tools programmers who live in the terminal and want AI editing folded into existing git workflows. It is open-source: you run it locally and bring your own model API keys. Tracked here because it appears in tools and pipeline contexts as a developer's AI-assisted coding option, reflecting the spread of agentic coding into production engineering.
39 mentions
Claude Codecode
Anthropic
Claude Code is Anthropic's terminal-based coding agent, applying long-context Claude reasoning across a whole project tree - reading, editing and running code from the command line with awareness of the repo and its tests. Its strength is sustained, multi-file work on real codebases rather than single snippets, which suits the kind of pipeline, tooling and automation engineering behind productions. In creative-tech it is relevant to TDs and tools programmers adopting agentic coding in their terminal workflows. It is proprietary and hosted, running on Claude models, with no self-hosting. Tracked here because AI-assisted development is increasingly named in creative-tech tools roles; it reflects the move from autocomplete toward agents that handle whole tasks across a project.
20 mentions
ComfyUIplatform
Comfy Org
ComfyUI is the node-based workflow editor that has become the de-facto pipeline tool for serious Stable Diffusion and Flux work. Instead of a one-shot prompt box it exposes the whole generation graph - models, samplers, ControlNets, LoRAs, upscalers, video nodes - as connectable nodes, so artists build reusable, reproducible pipelines and share them as workflow files. That control is exactly what production teams need: a fixed graph yields consistent output across a shot or a series. It is open-source and self-hostable, running locally on a workstation GPU, which keeps unreleased assets in-house. In creative-tech hiring, ComfyUI is shorthand for hands-on generative pipeline skill rather than casual prompt use.
15 mentions
Cursorcode
Anysphere
Cursor is an AI-first code editor forked from VS Code, built around deep codebase awareness - it indexes a project so its completion, chat and multi-file edits understand the whole repo, not just the open file. It became a favourite of developers for agentic refactors and describe-the-change-and-let-it-edit workflows. In creative-tech it is relevant to pipeline TDs, tools programmers and technical artists who write the glue code, shaders and automation behind productions; postings increasingly list AI-assisted coding fluency as a plus. It is proprietary with a free tier and no self-hosting, running on top of hosted models. Tracked here because it shows up in tools and pipeline job requirements, not because it touches art directly.
15 mentions
ChatGPTtext
OpenAI
ChatGPT is OpenAI's flagship assistant and the product that brought generative AI mainstream. In creative-tech postings it appears constantly - not as a craft tool but as a baseline expectation of AI literacy, and for everyday tasks like drafting documentation, writing and debugging pipeline scripts, summarising research and ideating. Its multimodal versions also handle image input and generation. It is fully hosted with a free tier and no self-hosting; control is conversational, with a tools ecosystem and an API for integration. Tracked here because it is the single most-named AI tool in job requirements across disciplines, used as a general productivity and reasoning aid rather than for producing final art, animation or VFX.
14 mentions
GitHub Copilotcode
GitHub / Microsoft
GitHub Copilot is the longest-running mainstream coding assistant, offering inline completion and chat across major editors (VS Code, JetBrains, Neovim) and inside GitHub itself. Backed by Microsoft and OpenAI models, it became the default first taste of AI pair-programming for many developers. In creative-tech it matters to the engineers behind productions - pipeline, tools and gameplay programmers, technical artists - and appears in postings as a baseline expectation of AI-assisted development. It is proprietary with no self-hosting, running on hosted models with enterprise controls for code privacy. Tracked here because creative-tech tooling roles increasingly name it as a working tool, reflecting how AI coding assistants have become standard equipment rather than a novelty.
13 mentions
Stable Diffusionimage-gen
Stability AI
Stable Diffusion is the open-weights text-to-image model that turned generative imaging into something studios can run on their own hardware. Because the weights are public it spawned a huge ecosystem - community checkpoints, LoRAs, ControlNet and ComfyUI node graphs - that lets artists fine-tune a look, lock a character, or drive generation from depth and pose maps. In production it appears in concept art, texture and matte ideation, and previs, where running locally keeps unreleased IP off third-party servers. The SDXL and SD3.5 generations improved prompt-following and text rendering, though many pipelines still favour older 1.5 checkpoints for the depth of community tooling built around them.
9 mentions
InvokeAIplatform
Invoke
InvokeAI is a production-oriented Stable Diffusion stack that wraps the open diffusion ecosystem in a polished canvas UI with unified model management, a robust inpainting and outpainting canvas, and team collaboration features. It targets studios that want the control of open-weights generation without assembling everything from raw nodes, and offers a commercial and enterprise tier alongside the open-source project. Being self-hostable, it keeps assets and unreleased IP on the studio's own infrastructure. In creative-tech work it appears in concept and production-art pipelines where a managed, multi-artist SD environment beats a single-user setup. Control comes through its canvas, model library and the underlying SD and Flux models it runs.
6 mentions
Geminitext
Google
Gemini is Google's multimodal model family (Pro, Advanced, 2.x), integrated across Workspace, Android and Vertex AI, and the front door to Google's image (Nano Banana) and video (Veo) generation. In creative-tech it appears as a general AI assistant for writing, research, scripting and analysis, and is especially relevant at organisations standardised on Google Cloud and Workspace. It is fully hosted with a free tier and no self-hosting; access is through the Gemini apps and Vertex AI for enterprise integration. Tracked here because it is one of the most-named general AI assistants in postings, signalling baseline AI literacy, and because it is the gateway to Google's generative image and video models that do touch creative pipelines.
5 mentions
Midjourneyimage-gen
Midjourney, Inc.
Midjourney is a proprietary text-to-image generator known for a polished, painterly default aesthetic that needs little prompt engineering to look good. It runs through Discord and a web app rather than a local install, so there is no self-hosting and no weights to fine-tune - control comes from prompts, style references and parameters. In creative-tech work it is most common in early concept and mood-board stages, where its strong sense of composition and lighting produces pitch-ready frames fast. The trade-off for that quality is a closed pipeline: output runs through Midjourney's servers, the look is recognisable enough that some studios treat it as ideation rather than final art, and integration into automated pipelines came late.
4 mentions
Runwayvideo-gen
Runway
Runway is a web-based video generation and editing suite whose Gen-2, Gen-3 and Gen-4 models helped define the text- and image-to-video category. Beyond raw generation it ships director-style controls - camera moves, motion brush, and tools like Act-One for driving a character's performance from a reference video - which is why it reads as a production tool rather than a toy. In creative-tech work it appears in previs, concept films, title and ident sequences, and quick effect shots, often as a faster alternative to full CG for short pieces. It is fully hosted with a free tier and no self-hosting, so footage runs through Runway's cloud - a consideration for studios handling unreleased IP.
4 mentions
Codeiumcode
Codeium / Windsurf
Codeium is a code completion and chat assistant available free across many IDEs, and the same company makes Windsurf, an agentic AI editor that can plan and execute multi-step changes across a codebase. The free, broad-IDE access made Codeium a popular alternative to Copilot, while Windsurf competes with Cursor in the agentic-editor space. In creative-tech it is relevant to pipeline, tools and gameplay engineers and technical artists writing production code. It is proprietary with a free tier and no self-hosting, running on hosted models with enterprise privacy options. Tracked here because it surfaces in tools and pipeline job requirements as part of the now-standard AI-assisted development toolkit rather than for any direct role in art or VFX.
4 mentions
RealityScan3d
Epic Games / Capturing Reality
RealityScan is the photogrammetry app from Epic Games' Capturing Reality team, the mobile-and-desktop lineage of RealityCapture - the high-end tool widely used to turn photo sets into dense, accurate 3D meshes. Its strength is fidelity: for hero props, environment scans and set pieces that must match reality, photogrammetry beats generative 3D on accuracy, and RealityCapture is a studio standard for that work. Tied into Epic's ecosystem, it has a clear path into Unreal Engine. It is proprietary with a free tier for the app, and the desktop RealityCapture runs locally on a workstation. In creative-tech hiring it signals VFX and virtual-production capture skill rather than generative AI per se.
3 mentions
Llamatext
Meta
Llama is Meta's open-weight LLM family (Llama 3.x, 4.x) and the dominant open base for fine-tuning and self-hosting - the foundation under a huge share of custom and on-prem AI deployments. For studios that need an LLM they fully control, Llama is usually the starting point: run it locally, fine-tune it on internal data, and keep everything in-house. In creative-tech it is relevant to engineering and tools teams building private AI features and pipelines. It is open-weight and self-hostable, with hosted access available through many providers. Tracked here because it appears in postings wherever self-hosted or fine-tuned LLM experience is wanted, representing the open-source backbone of the AI ecosystem rather than a consumer chatbot.
2 mentions
DaVinci Resolve AIvideo-gen
Blackmagic Design
DaVinci Resolve AI refers to the DaVinci Neural Engine - the set of AI features inside Blackmagic's Resolve, the editing, colour, VFX and audio suite widely used in film and post. Features like Magic Mask (roto without manual tracking), Voice Isolation, Smart Reframe, depth maps and AI relighting fold machine learning into a real production tool that editors and colourists already rely on. In creative-tech it is squarely a post-production tool, and the AI features speed up tasks - rotoscoping, audio cleanup, reframing - that used to be slow and manual. Resolve has a capable free version and runs locally on a workstation. Tracked here because postings name its AI features as practical, in-pipeline editing and finishing skills.
2 mentions
DALL-Eimage-gen
OpenAI
DALL-E is OpenAI's text-to-image model, now reached mainly through ChatGPT and the OpenAI API rather than as a standalone product. Its strengths are instruction-following and typography - it handles multi-part prompts and in-image text better than most diffusion models - which suits quick concept frames, storyboards and mock-ups that need legible signage or UI. Because it is fully hosted and content passes through OpenAI's moderation, it fits ideation and internal use more than locked-down production art. There is no self-hosting and no fine-tuning of the base model; control is through natural-language prompts and conversational edits inside ChatGPT, which lowers the barrier for non-specialists on a team.
1 mention
Imagenimage-gen
Google DeepMind
Imagen is Google DeepMind's text-to-image model family, reached through Vertex AI and integrated into the Gemini apps rather than sold as a standalone generator. Recent versions push photorealism, prompt adherence and improved text rendering, positioning it as Google's answer to the leading proprietary image models. For studios it is most relevant when they are already on Google Cloud - Vertex AI gives an enterprise path with the access controls and billing that production teams need. It is fully hosted with no self-hosting or weight access; control is through prompts and API parameters. In practice it appears in ideation and content pipelines, especially at organisations standardised on Google's cloud and Gemini tooling.
1 mention
NovelAIimage-gen
Anlatan
NovelAI is a subscription service combining an anime-focused image generator with an AI writing assistant, built around permissive content rules and strong character consistency. Its image models, tuned heavily on illustration and anime art, made it popular for character design, visual-novel assets and stylised concept work in that idiom, with dedicated tools for training and holding a character's look. It is fully hosted with no free tier and no self-hosting; control comes through prompts, model presets and its character tools. In creative-tech pipelines it is niche - most relevant to anime, games and visual-novel studios - rather than a general VFX or film tool, but within that lane it has a loyal following.
1 mention
Klingvideo-gen
Kuaishou
Kling is a text- and image-to-video model from Kuaishou, the Chinese short-video company, and quickly earned a reputation for high physical realism and relatively long, coherent clips. It handles human and object motion convincingly, which made it a frequent benchmark against Runway and Sora in the generative-video race. Studios reach for it on concept films, social content and quick effect tests where believable motion matters. It is fully hosted through Kling's web app and API with no self-hosting; footage runs on Kuaishou's servers, and access has at times been region-gated, which some Western studios weigh when handling client material. Control is through prompts, reference images and motion settings.
1 mention
Veovideo-gen
Google DeepMind
Veo is Google DeepMind's video generation model, reached through Vertex AI and the Gemini apps rather than as a standalone editor. It targets high-resolution, longer clips with strong prompt adherence and, in later versions, synchronised audio - positioning it as Google's flagship answer to Sora and Runway. For studios it is most relevant when already on Google Cloud, where Vertex AI provides the enterprise access controls and billing production teams need. It is fully hosted with no self-hosting or weight access; control is through prompts and API parameters. In practice it shows up in concept and content pipelines, especially at organisations standardised on Google's cloud and Gemini tooling.
1 mention
Higgsfieldvideo-gen
Higgsfield AI
Higgsfield is a video generator built camera-first: instead of hoping a prompt yields the right move, it offers a library of cinematic camera presets - dolly, crash zoom, orbit, bullet-time - applied to a generated or supplied shot. That focus makes it popular with filmmakers and social creators who want recognisable, motion-driven sequences fast. In creative-tech work it leans toward stylised concept pieces, trailers and motion tests rather than plate-quality VFX. It is fully hosted with a free tier and no self-hosting, so footage runs in its cloud. Control comes through the camera and motion presets, prompts and image inputs, trading fine frame-level control for quick, dramatic camera language.
1 mention
Topaz Labsupscale
Topaz Labs
Topaz Labs builds a suite of AI-powered upscalers and restorers - Gigapixel for stills, Video AI for footage and Photo AI for denoise and sharpening - that have become a quiet staple of VFX and post pipelines. Studios reach for them to rescue low-resolution plates, uprez and deinterlace archival footage, recover detail, and clean noise from high-ISO shoots without re-rendering. Unlike generative upscalers, Topaz aims to enhance the real source rather than invent detail, which is why finishing and restoration work trusts it. The models run locally on a workstation GPU rather than the cloud, keeping client footage in-house - a practical reason it appears in studio toolchains. It ships as desktop apps with per-product licensing.
1 mention
Hugging Faceplatform
Hugging Face
Hugging Face is the central hub of the open machine-learning world: the place where open-weights models, datasets and hosted demo apps (Spaces) are published, versioned and discovered. For creative-tech it is effectively where the open diffusion and video ecosystem lives - Stable Diffusion, Flux and many video and 3D model weights are distributed here, pulled via the Diffusers library or the Inference API. Studios use it to source models for local pipelines, host internal Spaces, and track what is new in open generation. It supports self-hosting (download the weights and run them anywhere) alongside a hosted Inference API, with organisations and access controls for teams. In hiring it signals fluency with the open-model supply chain behind in-house generative tooling.
1 mention
Luma AI3d
Luma
Luma AI is best known for photoreal scene capture - reconstructing a place or object as a NeRF or Gaussian Splat from ordinary phone video, no rig required. For VFX and virtual production that is a fast route to digital sets, reference environments and hero props captured on location and brought into the pipeline. It is the capture sibling to Luma's Dream Machine video model, so a team can scan and generate from the same vendor. It is fully hosted with a free tier and no self-hosting; capture runs on Luma's cloud, and output exports to common 3D and splat formats. In creative-tech work it shows up wherever real-world environments need to become usable 3D quickly rather than be modelled by hand.
1 mention
Spline3d
Spline Design
Spline is a browser-based 3D design tool popular for interactive web 3D - product scenes, landing-page animations and UI in three dimensions - now with AI features for generating materials, environments and basic text-to-3D. Its appeal is approachability: it brings real-time 3D design to people who are not Blender or Maya users, with collaboration and easy web export. In creative-tech terms it sits closer to web and product design and motion than to VFX or game pipelines, though concept and design teams use it for quick 3D mock-ups and presentations. It is fully hosted in the browser with a free tier and no self-hosting; the AI features assist within its design canvas rather than replacing a full DCC.
1 mention
RADiCALanimation
RADiCAL Inc.
RADiCAL is an AI motion-capture system that extracts 3D body motion from a single ordinary camera - no suit, markers or multi-cam rig - and outputs to common 3D pipelines. The single-camera approach trades some accuracy for extreme accessibility: anyone with a phone can capture rough body animation for previs, prototyping and indie work. It exports skeleton data for cleanup and retargeting in standard DCCs and engines. It is fully hosted with a free tier and cloud processing, no self-hosting. In creative-tech terms it sits in accessible mocap alongside DeepMotion and Plask, most useful where convenience and cost matter more than the precision of a marker-based stage - so previs and prototyping rather than hero performance capture.
1 mention
Nano Bananaimage-gen
Google
Nano Banana is the community nickname for Google's image generation and editing model delivered inside Gemini (Gemini 2.5 Flash Image), which earned a strong reputation for high-quality text-to-image and especially for natural-language image editing - changing parts of an image just by describing the change. It handles consistent edits and compositing instructions well, making it useful for quick concept iteration and mock-ups. It is fully hosted through Gemini with a free tier and no self-hosting or weight access; control is conversational, through prompts and follow-up edits. In creative-tech pipelines it fits ideation and rapid editing rather than locked-down production art, and reflects Google folding image generation directly into its assistant.
1 mention
CapCut AIvideo-gen
ByteDance
CapCut AI refers to the AI features inside CapCut, ByteDance's fast-growing video editor - auto-cut, AI script-to-video, background removal, captions and voice cloning - that turn rough footage into finished short-form content quickly. It became the default editor for a huge wave of social and creator video, bringing AI editing to a mass, non-specialist audience. In creative-tech terms it sits in social, marketing and content production rather than VFX or film post, but it appears in postings for content-team roles. It is desktop and cloud with a free tier and no self-hosting. Tracked here because it is one of the most widely used consumer AI editors, and content roles increasingly name it - noting that, as a ByteDance product, some organisations weigh data handling.
1 mention
Adobe Fireflyimage-gen
Adobe
Adobe Firefly is Adobe's family of generative models trained on licensed Adobe Stock and public-domain content, which is its main selling point: output is marketed as commercially safe and indemnified, easing the legal questions studios have about generative art. Rather than a standalone destination, Firefly is embedded across Photoshop (Generative Fill), Illustrator and Express, so it slots into the workflows artists already use instead of replacing them. In production it is common for cleanup, extending plates, generating background elements and quick variations directly inside those tools. It is fully hosted by Adobe with a free tier tied to generation credits, and leans on rights-cleared training rather than raw open-ended novelty.
no mentions yet
Leonardo AIimage-gen
Leonardo Interactive
Leonardo AI is a hosted image generator aimed at game and entertainment art teams, layering fine-tuned models, trained elements and a project-based UI on top of diffusion. Its appeal is workflow rather than a single signature look: reusable trained styles, character and asset consistency tools, and game-art-oriented presets for concept work, props and environment ideation. That makes it a frequent pick for studios that want repeatable production art instead of one-off images. It runs in the browser with a credit-based free tier, and was acquired by Canva, which pulls it further toward design and content teams. There is no self-hosting; control comes through trained models, image guidance and prompt tuning inside the platform.
no mentions yet
Ideogramimage-gen
Ideogram, Inc.
Ideogram is a hosted text-to-image generator best known for rendering legible in-image text - words, logos and typographic layouts - far more reliably than most diffusion models, which historically garble lettering. That niche makes it a practical tool for poster comps, title treatments, mock packaging and any concept frame where readable type matters. Beyond text it produces clean, design-oriented imagery with style presets and aspect controls. It is fully hosted in the browser with a free tier; there is no self-hosting or weight fine-tuning, so control is through prompts, style settings and reference images. In creative-tech pipelines it tends to appear in graphic-design-adjacent ideation rather than VFX plate work.
no mentions yet
Recraftimage-gen
Recraft AI
Recraft is a hosted image generator aimed at designers, with a standout ability to produce true vector output (SVG) and to hold a consistent brand style across a set of images. That makes it less about cinematic frames and more about logos, icons, illustrations and on-brand graphic assets that need to scale cleanly and stay visually coherent. It runs in the browser with a free tier; there is no self-hosting, and control comes through style definitions, prompts and its design-focused canvas. In creative-tech work it sits closer to graphic design, marketing art and UI ideation than to VFX or game-engine assets, though concept teams use it for clean stylised reference.
no mentions yet
Kreaplatform
Krea AI
Krea is a hosted creative canvas built around real-time generation: as you sketch or adjust a prompt the image updates live, which feels more like interactive art direction than batch prompting. It bundles image generation, real-time enhance and upscale, and video tools, and aggregates several underlying models so users can switch engines without leaving the canvas. The real-time loop suits fast concept exploration, style tests, and turning rough blocking into rendered frames within a session. It runs in the browser with a free tier; there is no self-hosting. In creative-tech pipelines it shows up in early ideation and look-development, where speed of iteration matters more than final-frame control.
no mentions yet
Playgroundimage-gen
Playground AI
Playground is a hosted text-to-image generator aimed at a broad creative audience, pairing its own tuned models with curated style filters and a social, Pinterest-like discovery feed for remixing others' work. The emphasis is approachability: presets and an editing canvas lower the barrier for users who do not want to prompt-engineer from scratch. It runs in the browser with a generous free tier; there is no self-hosting or fine-tuning of the base model. In creative-tech work it leans toward quick concept and mood imagery and personal exploration rather than controlled production assets, though its style presets can be a fast way to lock a consistent look for a pitch.
no mentions yet
Civitaiplatform
Civitai
Civitai is not a generator but the dominant community hub for open-weights image models - the place where Stable Diffusion and Flux checkpoints, LoRAs, ControlNets and embeddings are shared, rated and discovered. For any studio running a local diffusion pipeline it is effectively the app store: an artist pulls a trained style or character LoRA from Civitai, drops it into ComfyUI or Automatic1111, and reproduces a look without training from scratch. It also hosts on-site generation and bounties. Its relevance to creative-tech hiring is as connective tissue - the reason a 'Stable Diffusion' requirement usually implies fluency with the community tooling Civitai aggregates - rather than as a product studios license.
no mentions yet
Soravideo-gen
OpenAI
Sora is OpenAI's text-to-video model, notable for high visual fidelity, longer clip durations and strong character and scene continuity compared with earlier generators. Reached through OpenAI's apps and API rather than as a standalone craft tool, it targets quick, polished sequences from a prompt or reference image. In creative-tech pipelines it sits in concept films, previs and pitch material, where convincing motion and consistency sell an idea before any CG is built. It is fully hosted with no self-hosting and no weights to fine-tune; content passes through OpenAI's moderation, so it fits ideation and internal use more naturally than locked-down final shots on confidential projects.
no mentions yet
Pikavideo-gen
Pika Labs
Pika is a web-based video generator focused on short, stylised clips and playful effects rather than photoreal long-form footage. It became popular for its accessible interface and signature 'Pikaffects' - quick transformations and motion effects applied to an image or clip - which suit social content, idents and lightweight motion graphics. In creative-tech work it tends toward ideation, mood pieces and stylised inserts more than plate-quality VFX. It is fully hosted with a free tier and no self-hosting, so generation runs in Pika's cloud. Control comes through prompts, image inputs and effect presets, keeping the barrier low for non-specialists who need a quick animated beat.
no mentions yet
Luma Dream Machinevideo-gen
Luma AI
Dream Machine is Luma AI's text- and image-to-video model, the generative-video sibling to the company's NeRF and Gaussian-Splatting capture stack. It produces smooth, naturalistic motion and camera moves from a prompt or still, and added keyframe-style controls for guiding the start and end of a shot. In creative-tech pipelines it appears in previs, concept films and quick establishing shots, with the bonus that Luma's capture tools sit alongside it for teams blending generated and scanned content. It is fully hosted with a free tier and no self-hosting, so footage runs on Luma's cloud. Control is through prompts, reference frames and the keyframe interface.
no mentions yet
Seedancevideo-gen
ByteDance
Seedance is ByteDance's text- and image-to-video model, built for high fidelity with a particular strength in character continuity and complex, choreographed motion - hence the name. It competes at the top of the generative-video field on realism and the ability to hold a subject consistent across a clip. Studios use it for concept films, social content and motion tests where believable, sustained movement matters. It is fully hosted with no self-hosting; footage runs on ByteDance's infrastructure, which - as with other Chinese-developed models - some Western studios weigh when handling confidential client material. Control is through prompts, reference images and motion guidance.
no mentions yet
Wanvideo-gen
Alibaba (Wanxiang)
Wan is Alibaba's Wanxiang video and image generation family, and its open-weight Wan 2.1 and 2.2 releases became favourites in the Stable Diffusion community for running generative video locally. Because the weights are public it plugs into ComfyUI and lets studios keep footage in-house and build custom pipelines - a meaningful difference from the fully hosted competitors. It covers text- and image-to-video with steadily improving motion and resolution, and also offers a hosted path through Alibaba's cloud. In creative-tech work it appears wherever local control and data privacy matter for previs, concept and effect tests, and it anchors a growing open-weight video ecosystem alongside Hunyuan Video.
no mentions yet
Hailuovideo-gen
MiniMax
Hailuo is MiniMax's web video generator (the model line is also called MiniMax-Video), known for strong realism in human and animal motion and expressive facial performance. It became a popular pick for character-driven short clips and social content where believable bodies and faces carry the shot. In creative-tech pipelines it sits in concept work, motion tests and stylised inserts rather than plate-quality VFX. It is fully hosted with a free tier and no self-hosting, so footage runs on MiniMax's servers - the usual data-handling consideration for confidential projects. Control is through prompts, reference images and motion settings, with an accessible interface aimed at a broad creator base.
no mentions yet
Hunyuan Videovideo-gen
Tencent
Hunyuan Video is Tencent's open-weight video diffusion model and one of the first major open-weight releases to approach Sora-class quality, which made it a landmark for the local generative-video community. Because the weights are public it runs in ComfyUI and on a studio's own GPUs, so footage and unreleased IP need never leave the building - the key reason it appears in privacy-sensitive pipelines. It covers text- and image-to-video with strong motion and a fast-growing ecosystem of community fine-tunes and LoRAs. In creative-tech work it anchors local previs, concept and effect-test workflows alongside Alibaba's Wan, and signals a studio's investment in self-hosted generative video.
no mentions yet
LTX Videovideo-gen
Lightricks
LTX Video is Lightricks' open-weight video model, designed for speed - it generates short clips fast enough to feel near-real-time on capable hardware, trading some maximum fidelity for rapid iteration. Being open-weight, it runs locally and in ComfyUI, so studios can keep generation in-house and fold it into custom pipelines. The quick turnaround suits ideation, animatics and high-volume style tests where waiting on a cloud queue would break the creative loop. It covers text- and image-to-video. In creative-tech work it appears wherever fast local iteration matters more than the absolute top of the quality curve, and it broadened the open-weight video field beyond the larger, slower models.
no mentions yet
Stable Video Diffusionvideo-gen
Stability AI
Stable Video Diffusion is Stability AI's open-weight image-to-video (and later text-to-video) model, one of the earliest open options for local video generation. Though newer open-weight models have surpassed its fidelity and length, SVD remains a familiar entry point: it runs in ComfyUI and on a workstation GPU, so studios can experiment with generated motion without sending footage to a cloud service. In creative-tech work it shows up in early experiments, animatics and short image-to-video inserts, and as a teaching and reference model in the open ecosystem. Control is through the conditioning image, motion parameters and the surrounding ComfyUI graph rather than a polished hosted UI.
no mentions yet
HeyGenvideo-gen
HeyGen
HeyGen is an avatar video platform built around realistic talking presenters and high-quality lip-sync, including the ability to re-voice and re-lip footage into other languages. Its core use is localisation, training, explainer and product-demo video - turning a script or an existing clip into a polished spokesperson piece without a shoot. In creative-tech work it sits in marketing, internal comms and content production rather than VFX or game art, though localisation teams use its translation and lip-sync on existing footage. It is fully hosted with a free tier and no self-hosting; control comes through avatar selection, script input and voice settings. Likeness and consent controls matter here, since the output is synthetic people.
no mentions yet
Hedravideo-gen
Hedra
Hedra is a character-focused video generator with a strength in expressive facial animation and lip-sync driven from an image plus audio - feed it a portrait and a voice track and it animates a performance. That makes it popular for talking-head clips, prototype animatics and quick character beats where a full facial-capture pipeline would be overkill. In creative-tech work it suits previs, pitch pieces and stylised character tests more than final shots. It is fully hosted with a free tier and no self-hosting, so media runs in its cloud. Control comes through the source image, audio and prompt; as with any likeness-driven tool, consent for the face and voice used is the practical constraint.
no mentions yet
D-IDvideo-gen
D-ID
D-ID is a long-standing talking-avatar service that animates a still photo into a speaking head, with strong photo-to-video lip-sync and a streaming API for real-time digital humans. It is widely used for explainer video, avatar presenters, customer-facing digital humans and localisation, turning a single image and a script into a talking clip. In creative-tech pipelines it sits in marketing, product and conversational-AI work rather than VFX or game art. It is fully hosted with a free tier and no self-hosting; control is through the source image, script or audio, and voice selection. Because it generates synthetic likenesses, consent and its content policies are the main guardrails on use.
no mentions yet
Synthesiavideo-gen
Synthesia
Synthesia is a studio-style avatar video platform and the de-facto standard for corporate-training and marketing AI presenters. From a script it generates a polished spokesperson video with a library of stock avatars, custom avatars and many languages - replacing a studio shoot for explainer, onboarding and L&D content. In creative-tech terms it sits squarely in enterprise content and communications rather than VFX, animation or game art, but it appears in postings as a named tool for content and marketing teams. It is fully hosted with no self-hosting; control is through avatar choice, script and template. Its enterprise focus means strong emphasis on consent, brand control and moderation of synthetic presenters.
no mentions yet
Figma Weaveplatform
Figma
Figma Weave is Figma's node-based AI canvas - effectively a ComfyUI analogue tuned for product, brand and design teams rather than VFX. It lets designers chain generation, editing and variation nodes to build repeatable workflows for mood boards, asset variants and on-brand imagery, all inside the Figma ecosystem they already work in. The appeal is bringing structured, reproducible generative workflows to people who live in design tools rather than ComfyUI graphs. It is fully hosted by Figma with a free tier and no self-hosting. In creative-tech work it sits in product design, branding and marketing art rather than film or game pipelines, and signals generative fluency within a design-team context.
no mentions yet
Magnificupscale
Magnific AI
Magnific is a generative upscaler that does more than sharpen - it hallucinates plausible new detail as it increases resolution, inventing texture, pores and micro-structure guided by a creativity slider and prompt. That makes it a finishing tool for concept art and stylised imagery, where invented detail is welcome, rather than a faithful restoration tool where accuracy to the source matters. Artists use it to take a rough or low-res generation to a high-res, portfolio-ready frame. It is fully hosted (and was acquired by Freepik) with no self-hosting, so images run in its cloud. The trade-off to understand is fidelity: it improves perceived detail by adding, not recovering, so it is unsuitable where output must match real source content.
no mentions yet
Automatic1111platform
AUTOMATIC1111 (community)
AUTOMATIC1111 (the Stable Diffusion WebUI) is the Gradio-based interface that, before ComfyUI, was the default way to run Stable Diffusion locally. It wraps the open diffusion ecosystem in a tabbed UI - txt2img, img2img, inpainting - plus a large extension system that adds ControlNet, LoRA and upscalers, so an artist can run SD on their own GPU without touching code. Its huge community library of extensions and scripts made it the reference point for hands-on local generation, and many tutorials and checkpoints still assume it. In creative-tech work it signals practical, self-hosted diffusion experience that keeps unreleased assets in-house. It is open-source and self-hostable; control comes through its UI, extensions and the underlying SD and Flux models it runs.
no mentions yet
Replicateplatform
Replicate
Replicate is a cloud platform for running open-source AI models through a simple API, turning a model into a hosted endpoint you call without managing GPUs. Its catalogue spans image, video, audio and language models pushed by the community and packaged with Cog, so a studio's tools team can wire generation into pipelines, web apps and internal tools over HTTP rather than standing up inference infrastructure. Pricing is pay-per-second of compute. In creative-tech work it sits on the engineering and pipeline side - how TDs and tools programmers integrate generative steps - rather than as an artist-facing canvas. There is no self-hosting; control is through its API, model versions and the parameters each model exposes.
no mentions yet
fal.aiplatform
fal
fal (fal.ai) is an inference platform built for speed, offering low-latency, serverless API access to generative media models - especially the Flux image family and fast video models. Its pitch is performance: optimised runtimes and streaming that make generation feel near-real-time, which suits interactive tools, demos and high-volume pipelines. Like Replicate it is developer-facing - tools and pipeline engineers reach for it to embed image and video generation behind an API without owning GPU infrastructure - rather than an artist canvas. It is fully hosted with a free allowance and no self-hosting; control is through its API, the hosted model endpoints and their parameters. In creative-tech it appears in tooling and prototyping contexts where generation speed and easy integration matter.
no mentions yet
Fooocusplatform
lllyasviel (community)
Fooocus is a streamlined Stable Diffusion interface from lllyasviel (the author of ControlNet), built to feel like Midjourney: one prompt box, strong defaults, and the fiddly settings hidden away. It runs SDXL locally and quietly handles the prompt engineering, sampler choices and refinements that a tool like AUTOMATIC1111 exposes, so a user gets high-quality images without learning the full pipeline. That makes it a fast on-ramp to local generation for artists who want results over control. It is open-source and self-hostable, running on a local GPU with a free, no-account workflow. In creative-tech terms it sits in quick concept and ideation rather than controlled production, most useful where ease and image quality matter more than fine-grained pipeline control.
no mentions yet
RunComfyplatform
RunComfy
RunComfy is a cloud host for ComfyUI: it runs the node-based workflow editor in the browser on rented GPUs, with models, custom nodes and dependencies pre-installed, so a team can build and run heavy generation graphs without a local GPU or a fiddly install. It targets artists and small studios who want ComfyUI's reproducible pipelines on demand - spin up a machine, load a workflow, generate, shut it down - paying for the compute they use. Similar services like RunDiffusion do the same for the AUTOMATIC1111 / SD WebUI world. There is no self-hosting, since the point is hosted GPUs; control is the full ComfyUI graph plus the platform's machine and model management. In creative-tech work it lowers the hardware barrier to serious generative pipelines.
no mentions yet
Wonder3D3d
Open-source (academic)
Wonder3D is an open-source, research-born model that turns a single image into a textured 3D mesh by first generating consistent multi-view images and then reconstructing geometry from them. As an academic release it is more a technique and reference implementation than a polished product - run from its GitHub repo, self-hosted on a GPU - which is exactly why it matters to studios building custom asset pipelines or evaluating single-image reconstruction. Output suits rapid greyboxing, reference props and ideation rather than final game-ready topology. In creative-tech hiring it signals R&D and pipeline-TD work: someone comfortable wiring an open model into a tool rather than clicking a hosted service.
no mentions yet
Tripo3d
VAST AI
Tripo is a hosted text- and image-to-3D service from VAST AI, known for fast generation and clean PBR-textured meshes - feed it a prompt or a reference image and get a usable model in seconds to minutes. It targets game and content pipelines that need quick props, set dressing and concept assets without a manual modelling pass, and exposes an API for batch and integration work. Output usually needs retopology before it is truly game-ready, so it sits in ideation and blockout more than final assets. It runs in the browser with a free tier and no self-hosting; control is through prompts, reference images and generation settings. It is one of the more capable consumer text-to-3D options.
no mentions yet
Meshy3d
Meshy Inc.
Meshy is a web-based text- and image-to-3D generator aimed squarely at game asset workflows, pairing generation with retopology and texture-editing tools so output lands closer to usable than raw mesh dumps. It is popular with indie and prototyping teams who need props, characters and environment pieces fast, and offers an API for pipeline integration. As with most generative 3D, results suit blockout, concept and prototype stages, with a cleanup pass expected before final use. It runs in the browser with a free tier and no self-hosting; control comes through prompts, reference images, style settings and its editing tools. It competes directly with Tripo and Rodin in the consumer text-to-3D space.
no mentions yet
Polycam3d
Polycam
Polycam is a phone-based capture app supporting photogrammetry, LiDAR and Gaussian Splatting, prized for how fast it turns a quick walk-around into a usable 3D scan. On set and on location it is a go-to for grabbing reference, set pieces and prop scans without hauling a rig, exporting to common 3D formats for the pipeline. It trades the ultimate accuracy of full photogrammetry suites for speed and convenience, which suits previs, reference and blockout more than hero-asset fidelity. It is fully hosted with a free tier; capture happens on the phone with cloud processing. In creative-tech work it appears wherever artists need real-world geometry captured quickly rather than modelled.
no mentions yet
Kaedim3d
Kaedim
Kaedim is a 2D-image-to-3D service aimed at game studios, distinctive for combining AI generation with a human-in-the-loop review step so the meshes it returns are closer to game-ready than pure-AI output. That hybrid model targets art teams that want concept-to-asset turnaround without sacrificing topology quality, and it is sold as a production service rather than a casual generator. It has drawn scrutiny over how much is AI versus manual, which is worth noting, but the pitch is reliable, usable assets. It is fully hosted with no self-hosting; control is through the input art and order specs. In creative-tech work it sits in game-art production pipelines rather than VFX or film.
no mentions yet
3DFY3d
3DFY.ai
3DFY is a text- and image-to-3D platform that emphasises clean topology and PBR materials, positioning itself for studios that care about mesh quality rather than just a fast blob. It targets scalable asset generation - producing libraries of props and objects to spec - and offers API access for pipeline and batch use. As with all generative 3D, output suits blockout, background and prototype assets, with hero work still expecting an artist pass. It is proprietary and hosted with no self-hosting; control comes through prompts, reference images and generation parameters. In creative-tech work it sits alongside Tripo, Meshy and Rodin as a contender for the game-and-content text-to-3D pipeline.
no mentions yet
Hunyuan3D3d
Tencent
Hunyuan3D is Tencent's open-weight text- and image-to-3D model, the 3D counterpart to its open video and image work. Because the weights are public it runs locally and integrates into ComfyUI, letting studios generate meshes on their own hardware and fold 3D generation into custom pipelines - the same in-house, IP-safe appeal as the open diffusion ecosystem. It produces textured geometry from a prompt or image, suited to blockout, concept and prototype assets with cleanup expected for final use. It is open-source and self-hostable. In creative-tech work it anchors local, self-hosted generative-3D experiments and signals a studio investing in open 3D tooling rather than a hosted service.
no mentions yet
Rodin3d
DeemosTech
Rodin is DeemosTech's photo- and text-to-3D generator focused on high-detail characters with PBR output, and it became one of the standout indie 3D tools of 2024-25 for the quality of its meshes. It targets character and creature work where generative 3D usually struggles, producing geometry detailed enough to be a strong starting point for game and concept pipelines - though, like all generative 3D, a retopology and cleanup pass is expected for final assets. It is hosted with a free tier and no self-hosting; control is through reference images, prompts and generation settings. In creative-tech work it sits in character art and rapid asset ideation, competing with Tripo and Meshy at the higher-detail end.
no mentions yet
ElevenLabsvoice
ElevenLabs
ElevenLabs is the leading AI voice platform, known for natural-sounding text-to-speech and high-quality voice cloning across many languages. In creative-tech work it shows up in game prototyping (temp and even shipped character VO), localisation and dubbing, audiobook and podcast production, and any pipeline that needs voice without booking a booth. Its API and dubbing tools made it a default for studios automating dialogue at scale. It is fully hosted with a free tier and no self-hosting, so audio runs in its cloud. The serious consideration is consent and rights: cloning a real voice requires permission, and studios using it for character VO weigh union, likeness and disclosure rules. Control is through voice selection, cloning and fine-grained delivery settings.
no mentions yet
Sunomusic
Suno
Suno is a text-to-music generator that produces full songs - instrumentation and sung vocals with lyrics - from a short prompt, and it became the most talked-about AI music tool for how complete its output feels. In creative-tech work it is used for temp scores, song sketches, idents and placeholder music where licensing a track or booking a composer would be slow or costly. It is fully hosted with a free tier and no self-hosting; control is through prompts, style tags and lyric input. The hard consideration is rights: AI music sits in unsettled legal territory around training data and commercial use, so studios tend to treat Suno output as scratch and temp rather than cleared final music, pending clarity on licensing.
no mentions yet
Udiomusic
Uncharted Labs
Udio is a generative music platform similar in spirit to Suno, producing songs with vocals from text prompts, with a focus on audio quality, extended-form composition and more granular, stem-style editing of sections. It suits temp scores, song sketches and music placeholders for content, trailers and prototypes where a quick original-sounding track is needed. It is fully hosted with a free tier and no self-hosting; control is through prompts, style guidance and section editing. As with all generative music, the open question is rights and training-data provenance, so studios generally treat its output as temp and reference rather than cleared final music. It competes head-to-head with Suno at the top of the AI-music space.
no mentions yet
Murfvoice
Murf AI
Murf is a studio-grade text-to-speech platform offering a large library of natural voices across languages, aimed at explainer, e-learning, corporate narration and presentation audio. Its pitch is a complete production workflow - script, voice, timing, background music and sync - rather than just a raw TTS engine, which suits content and marketing teams more than character VO. In creative-tech terms it sits in corporate and instructional content rather than games or film, but it appears in postings as a named narration tool. It is fully hosted with a free tier and no self-hosting; control is through voice choice, pronunciation and emphasis controls. As with all synthetic voice, its licensed stock voices sidestep the consent issues of cloning a real person.
no mentions yet
Resemblevoice
Resemble AI
Resemble AI is a voice cloning and text-to-speech platform built around an API, used for character voice work, localisation and real-time voice in games and interactive media. Its strengths are fast voice cloning, speech-to-speech conversion and tooling for integrating generated voice into pipelines and engines, plus features for detecting AI-generated audio. That developer focus makes it a fit for studios embedding dynamic dialogue rather than one-off narration. It is fully hosted with no self-hosting; control is through cloned or stock voices and its API. The central constraint, as with any cloning tool, is consent and rights for the source voice - a live issue for game and film studios navigating performer and union agreements.
no mentions yet
Descriptvoice
Descript
Descript is an audio and video editor with a defining trick: it transcribes your recording and lets you edit the media by editing the text, so deleting a word removes it from the audio. It bundles AI features like Overdub voice cloning, filler-word removal, studio-sound cleanup and AI eye-contact correction, which made it a staple for podcast, YouTube and explainer production. In creative-tech terms it sits in content and post for spoken-word media rather than VFX or game art, but it appears in postings as a named production tool. It is fully hosted with a free tier and no self-hosting. Overdub's voice cloning carries the usual consent requirement, and the transcript-driven workflow is its main draw for fast dialogue editing.
no mentions yet
Adobe Podcastvoice
Adobe
Adobe Podcast is Adobe's audio AI suite, best known for Enhance Speech - a one-click tool that strips room noise and reverb and makes a rough recording sound close to a studio mic, widely regarded as the cleanest free dialogue de-noise available. For creative-tech and post teams it is a fast fix for location audio, remote interviews and scratch dialogue before a proper mix, and it also includes Mic Check and recording tools. It is fully hosted by Adobe with a free tier and no self-hosting; processing runs in the browser. In pipelines it sits in audio cleanup and content production rather than music or character VO, and its appeal is the quality-to-effort ratio: minimal setup, strong restoration on speech.
no mentions yet
Krispvoice
Krisp
Krisp provides real-time noise and echo cancellation for calls and recordings, sitting between the microphone and the app to strip background sound, keystrokes and cross-talk on the fly. It is widely deployed across remote production and post teams for clean review calls, remote sessions and recordings, and adds meeting transcription and notes. In creative-tech terms it is infrastructure rather than a creative tool - it does not generate anything - but it appears in postings as a named utility for distributed teams. It is fully hosted with a free tier; processing runs locally in real time on the device. Its value is reliability in live communication rather than offline restoration, complementing rather than replacing a proper audio cleanup pass.
no mentions yet
Cascadeuranimation
Nekki
Cascadeur is a character animation tool built around physics: instead of pure hand-keyframing, it uses AI-assisted auto-posing and physics simulation to help animators block believable action, balance and secondary motion fast. It is aimed at game and indie animation teams who want physically plausible movement without a full mocap pipeline, and exports to standard formats for Maya, Blender and engines. In creative-tech work it sits in keyframe animation and previs, accelerating the blocking and physics-correction stages rather than replacing an animator. It is proprietary with a free tier and runs locally as a desktop app. Its niche is the middle ground between manual keyframing and motion capture - AI as an assist, not an autopilot.
no mentions yet
Move AIanimation
Move AI
Move AI is a markerless motion-capture system that turns ordinary video - even multi-phone setups - into clean 3D skeleton data (FBX) without suits, markers or a dedicated stage. That dramatically lowers the cost of mocap, which is why it appears in game, previs and indie pipelines that could never justify an optical capture volume. Output drops into Maya, Blender, MotionBuilder and engines for cleanup and retargeting. It is fully hosted (cloud processing of the footage) with no self-hosting; control is through the capture setup and its processing tools. In creative-tech hiring it signals modern, accessible mocap workflows, and it is among the most credible markerless options for production-quality body motion.
no mentions yet
DeepMotionanimation
DeepMotion
DeepMotion's Animate 3D is a markerless motion-capture service that converts single-camera video into 3D animation, popular in indie game, VTuber and prototyping circles for its accessibility and features like face and hand tracking. Upload a clip and get retargetable skeleton data for standard DCCs and engines. Like other single-camera tools it favours convenience over the precision of an optical volume, so it suits previs, prototyping and content rather than hero performance capture. It is fully hosted with a free tier and cloud processing, no self-hosting; control is through the input video and its retargeting options. In creative-tech work it sits in the accessible-mocap tier alongside RADiCAL and Plask.
no mentions yet
Plaskanimation
Plask
Plask is a web-based AI motion-capture tool that extracts animation from video in the browser and integrates with major DCCs and game engines, pitched as an all-in-one animate-and-edit workspace. Its appeal is doing markerless mocap plus light animation editing without installing a heavy pipeline, which suits indie teams, prototyping and content creators. As with other single-source mocap, it trades precision for accessibility, landing in previs and prototype work more than hero capture. It is fully hosted with a free tier and no self-hosting; control is through the input video and its in-browser editing and export tools. In creative-tech terms it sits with RADiCAL and DeepMotion in the accessible video-to-animation tier.
no mentions yet
Rokokoanimation
Rokoko
Rokoko is a motion-capture brand spanning affordable inertial suits and gloves plus Vision, its markerless video-based mocap - and it is the most-mentioned indie mocap name in our postings. The combination lets teams scale from phone-video capture up to suit-based performance without changing ecosystems, with Studio software for cleanup, retargeting and live streaming into DCCs and engines. That breadth made it a default for indie games, virtual production and previs that need real mocap on a modest budget. The suits are hardware while Vision is software; control runs through Rokoko Studio. In creative-tech hiring it signals practical, production-grade mocap experience outside the high-end optical houses.
no mentions yet
Wonder Studioanimation
Autodesk / Wonder Dynamics
Wonder Studio (from Wonder Dynamics, acquired by Autodesk in 2024) is an AI VFX system that automatically detects a live actor in a shot and replaces them with a 3D CG character - handling tracking, body motion, lighting and compositing that would normally take a full VFX team. The pitch is collapsing a character-replacement pipeline into a largely automated cloud process, opening creature and character work to smaller teams and faster turnarounds. Output comes back with editable elements for finishing in standard tools. It is proprietary and hosted, now backed by Autodesk's pipeline ecosystem. In creative-tech terms it sits squarely in VFX and character work, and is one of the clearest examples of AI automating a traditionally labour-intensive shot pipeline.
no mentions yet
Mixamoanimation
Adobe
Mixamo is Adobe's free auto-rigging and animation service: upload a character mesh and it rigs it automatically, then lets you apply from a large library of ready-made motion clips. It has been a staple of character animation prototyping for years - the fastest way to get a rigged, walking, fighting character into Unreal, Unity or Blender for blockout and previs. It is not generative AI in the modern sense, but it appears in postings as a baseline pipeline tool. It is fully hosted and free with no self-hosting; control is through the auto-rigger and the motion library. In creative-tech work it sits in prototyping and previs rather than final animation, where bespoke rigs and motion usually take over, but as a starting point it is hard to beat.
no mentions yet
Cuebricanimation
Cuebric
Cuebric is an AI virtual-production tool that generates and parallax-projects 2.5D environments for LED-wall and set-extension work - turning a generated or supplied image into layered, camera-reactive backgrounds for in-camera VFX. It targets the virtual-production niche specifically: building immersive backdrops fast for stages that would otherwise need full CG environments or location shoots. Output is structured for compositing and stage pipelines rather than a flat render. It is proprietary and hosted, aimed at film and episodic production teams. In creative-tech terms it sits in virtual production and VFX, and is a clear example of generative imagery adapted to a very specific, high-value production workflow rather than general image creation.
no mentions yet
Devincode
Cognition AI
Devin is an autonomous software-engineering agent from Cognition that runs in its own sandboxed environment - planning, writing, testing and iterating on code with minimal supervision, aiming to complete whole tickets rather than just suggest lines. It generated heavy debate about how capable autonomous coding agents really are, and sits at the ambitious end of the AI-developer spectrum. In creative-tech it is most relevant to the engineering side - pipeline and tools teams experimenting with delegating routine implementation. It is proprietary and hosted with no self-hosting. Tracked here because it appears in forward-looking tools and engineering discussions, as a marker of where autonomous coding is heading rather than an established daily-driver for most teams.
no mentions yet
Mistraltext
Mistral AI
Mistral is a French AI lab notable for releasing genuinely open-weight models (Mistral, Mixtral, Codestral) alongside hosted products (Le Chat, Mistral Large). The open releases made it a favourite for teams that want to self-host an LLM - keeping data in-house and avoiding per-token API costs - which matters for studios wary of sending material to a third-party cloud. In creative-tech it is relevant to engineering and tools teams building internal AI features on a model they control. It offers both self-hosting (open weights) and a hosted API. Tracked here because it appears in postings where on-prem or private LLM deployment is a requirement, representing the open-weight, European alternative to the big US proprietary assistants.
no mentions yet
DeepSeektext
DeepSeek
DeepSeek is a Chinese AI lab whose open-weight models (DeepSeek V3 and the R1 reasoning model) drew global attention for strong performance at a fraction of the usual API cost, briefly upending assumptions about how expensive frontier-level reasoning had to be. Being open-weight, it can be self-hosted, which appeals to teams wanting capable reasoning on their own infrastructure. In creative-tech it is relevant to engineering and tools work, and to cost-sensitive internal AI features. It offers self-hosting (open weights) and a low-cost hosted API. Tracked here because it appears in postings and tooling discussions as a notable open-weight option, though some Western organisations weigh data-handling and provenance when adopting Chinese-developed models.
no mentions yet
Groktext
xAI
Grok is xAI's chat assistant, available through X (Twitter) and a dedicated app, positioned around real-time knowledge from the X firehose and a less filtered conversational style. It competes with the major assistants on general reasoning while leaning on its live-data access as a differentiator. In creative-tech it appears as one more named general AI assistant for research, writing and ideation rather than a craft tool, and is most visible to teams already in the X ecosystem. It is proprietary and hosted with a free tier and no self-hosting. Tracked here because it shows up in postings among the general-purpose assistants that signal AI literacy, not for any direct role in art, animation or VFX production.
no mentions yet
Perplexitytext
Perplexity AI
Perplexity is a search-grounded answer engine: rather than answering purely from a model's memory, it retrieves current web sources and cites them, which makes it a popular drop-in for research that needs traceable references. It can route across several underlying LLMs. In creative-tech it is used for fast, citable research - tech evaluations, market and reference gathering, competitive scans - more than for production work. It is fully hosted with a free tier and no self-hosting; control is conversational with source controls. Tracked here because it appears in postings as a named research tool, reflecting how citation-grounded AI search has become part of the everyday knowledge-work toolkit around productions rather than a creative instrument itself.
no mentions yet
NotebookLMtext
Google
NotebookLM is Google's source-grounded reasoning notebook: you upload documents and it answers strictly from them, with citations back to the source, plus a popular feature that turns your material into a conversational audio overview. The strict grounding makes it useful where staying inside a defined corpus matters - onboarding from a documentation set, summarising a bible or spec, or studying reference material. In creative-tech it is a knowledge and research aid rather than a production tool, handy for teams digesting large internal document sets. It is fully hosted with a free tier and no self-hosting. Tracked here because it appears in postings as a named productivity tool, reflecting the spread of document-grounded AI into everyday creative-tech knowledge work.
no mentions yet
Qwentext
Alibaba
Qwen is Alibaba's open-weight LLM family (Qwen 2, 2.5, 3), respected for strong code and math performance at competitive open-weight scale, with broad multilingual support. As open weights it can be self-hosted, appealing to teams that want capable models on their own infrastructure without per-token costs. In creative-tech it is relevant to engineering and tools work and to internal AI features built on a controllable model. It offers self-hosting (open weights) and hosted access. Tracked here because it appears in postings and tooling discussions as a leading open-weight option alongside Llama, Mistral and DeepSeek, though - as with other Chinese-developed models - some organisations weigh data-handling and provenance before adopting it.
no mentions yet
Adobe Senseimulti
Adobe
Adobe Sensei is Adobe's underlying AI and machine-learning platform - the engine powering features across Creative Cloud rather than a product you open directly. It drives Photoshop's Generative Fill and content-aware tools, Premiere's AI editing and audio features, Lightroom's denoise and masking, and more. In creative-tech it is everywhere by proxy: artists and editors use Sensei-powered features daily inside the Adobe tools that anchor most pipelines, even if they never say its name. It is proprietary and hosted within Adobe's ecosystem. Tracked here because postings reference Adobe's AI capabilities as a baseline, and because Sensei represents how deeply AI is now embedded in the standard creative toolset rather than bolted on as a separate app.
no mentions yet
How we measure - regex-detected tool names in posting text (Midjourney, Stable Diffusion, Runway, ComfyUI, ElevenLabs, Cursor, Claude, ChatGPT, Copilot, Gemini, Veo, Sora, Kling, etc., 89 tools in the dictionary).
A mention can be a required tool, a "nice to have", or context-of-work. We don't yet distinguish these levels - that's coming.
What we miss - tools used internally but not named in the posting, and any AI tool we haven't added to the dictionary yet. The list grows with each weekly review.
▼Roadmap · what's coming1 feature pending
~ 6 months External signals
Cross-reference HuggingFace trending models + GitHub topic stars to spot tools before they hit job postings. Lead indicator for what'll be required next year.
ongoing Dictionary expansion
89 tools tracked today (see method note above). We review the parser weekly and add new tools as the market introduces them.