# ReelStack

> The AI video lab: repos, skills, workflows and prompts, tested by people who ship video. Studio Hall's in-the-trenches resource for AI video: repos, agent skills, MCPs, ComfyUI workflows, pipelines and prompt techniques. Posts marked "Spotted, not yet tested" are research summaries, not test results. Tool scores are preliminary editorial score, pending test renders.

## Key pages
- [The Lab](https://thereelstack.com/lab): short posts on the latest finds
- [Skills & repos directory](https://thereelstack.com/skills): 37 skills, MCPs, repos and ComfyUI workflows, plus a request-a-skill form
- [Model Picker](https://thereelstack.com/model-picker): style + needs → sourced model recommendations
- [Workflows & pipelines](https://thereelstack.com/workflows): 12 step-by-step pipelines and a guide to turning a workflow into an agent skill
- [Prompting](https://thereelstack.com/prompts)
- [Creators to follow](https://thereelstack.com/creators)
- [Tool comparison](https://thereelstack.com/compare)
- [Methodology](https://thereelstack.com/methodology)

## Lab posts
- [HyperFrames: HeyGen's bet that agents should write video as HTML](https://thereelstack.com/lab/hyperframes-html-to-video-for-agents): An Apache-2.0 renderer that turns plain HTML, CSS and GSAP into frame-exact MP4, with a skill router that picks the workflow for you. (spotted, not yet tested; source https://hyperframes.heygen.com/)
- [claude-motion teaches an agent taste, then makes it check its own frames](https://thereelstack.com/lab/claude-motion-taste-and-self-review): Motion rules with actual numbers, a contact-sheet review loop, and a sound engine written in plain Python. (spotted, not yet tested; source https://github.com/whaleyxbt/claude-motion)
- [Free video from code: the canvas + Playwright + FFmpeg pipeline](https://thereelstack.com/lab/animate-skill-js-canvas-playwright-ffmpeg): No Remotion and no generator credits: Claude writes a function of time, a headless browser draws each frame, FFmpeg stitches them, and the agent checks its own frames. (based on the creator's video, not yet tested by us; source https://www.youtube.com/watch?v=rscb1DgJtNg)
- [Kevin Ngo's JavaScript-drawn films, and the 2D-to-Blender handoff](https://thereelstack.com/lab/kevin-ngo-js-drawn-films): Short story films where the model draws every frame in code, and a neat trick for turning a 2D cut into a 3D one. (spotted, not yet tested; source https://x.com/kevin_t_ngo)
- [Higgsfield inside Claude: MCP, Soul ID and Katana](https://thereelstack.com/lab/higgsfield-mcp-katana-in-claude): One connector URL turns Claude into a front end for Higgsfield's models, characters and audio. Katana adds reference-driven edits. (spotted, not yet tested; source https://x.com/higgsfield/status/2110378916963876908)
- [AI UGC: spend half your time on the starting frame](https://thereelstack.com/lab/starting-frame-is-half-the-job): The cheapest quality upgrade in AI UGC happens before you press generate. (spotted, not yet tested; source https://x.com/maxxmalist/status/2110346820341916028)
- [The 15-repo motion design stack, sorted by job](https://thereelstack.com/lab/fifteen-repos-motion-studio): A viral list of motion-design repos, which we checked one by one and grouped by what you're actually trying to ship. (spotted, not yet tested; source https://x.com/Mnilax/status/2109314840057565345)
- [Seedance draft mode, plus a review gate worth copying](https://thereelstack.com/lab/genjutsu-open-workflow-subject-replacement): Two cost-control habits from Sirio's video: iterate in 480p draft mode and finalise once in 1080p, and refuse to spend until a human has approved the inputs. (based on the creator's video, not yet tested by us; source https://www.youtube.com/watch?v=d6spCMcfkew)
- [What 418 shared prompts teach about one-shot code video](https://thereelstack.com/lab/opus-video-prompt-gallery-patterns): Skillry collects creators' prompts and remakes each video side by side. The prompts that work share a structure. (spotted, not yet tested; source https://skillry.dev/ai-videos/opus-5-5)
- [Pull a brand from a URL, then animate it: Jack Roberts' Claude Design levels](https://thereelstack.com/lab/brand-extraction-to-motion-library): Scrape the real logo, colours and fonts with Firecrawl first, so the agent animates the brand instead of inventing one. (based on the creator's video, not yet tested by us; source https://www.youtube.com/watch?v=HOXrLsVqinY)
- [Wan 3.0 in ComfyUI: reference tags for images, video and audio](https://thereelstack.com/lab/wan-3-reference-to-video-comfyui): Official templates for text-to-video, image-to-video and reference-to-video, with clips up to 30 seconds and synced audio. (spotted, not yet tested; source https://docs.comfy.org/tutorials/partner-nodes/wan/wan3-0)
- [An open-source audio stack for video: music, foley, lip-sync, timing](https://thereelstack.com/lab/open-music-and-audio-sync-stack): Five repos that cover most of the sound side of an AI video pipeline without a subscription. (spotted, not yet tested; source https://github.com/ace-step/ACE-Step-1.5)
- [Google Flow and Nano Banana 2.1: a five-prompt consistency routine](https://thereelstack.com/lab/google-flow-nano-banana-consistency): Create the subject, move it twice, edit one detail, lock a product, then animate the best still. Plus the limits Google itself lists. (based on the creator's video, not yet tested by us; source https://www.youtube.com/watch?v=Sdx8mQ2LW7A)
- [Six motion styles clients pay for, according to Higgsfield](https://thereelstack.com/lab/six-paid-motion-styles): A vendor's map of sellable formats, with the workflow tips worth keeping: storyboard first, HTML design sheets, and a credit tally that includes failed takes. (based on the creator's video, not yet tested by us; source https://www.youtube.com/watch?v=2OwMjg5As2g)
- [A two-layer launch video: code motion over a Seedance plate](https://thereelstack.com/lab/two-layer-launch-video-opus-higgsfield): Recreate a reference launch video's style for your own product, then layer that motion over cinematic footage built in a strict four-step order. (based on the creator's video, not yet tested by us; source https://www.youtube.com/watch?v=J4oygRI6mN0)
- [Ten GitHub repos from one creator's roundup, checked where we could](https://thereelstack.com/lab/ten-github-repos-jack-roberts-oct-2026): An open-source 'Adobe suite', a second Photoshop rebuild, an agent browser, a component MCP and more. We've linked only what we could confirm. (based on the creator's video, not yet tested by us; source https://www.youtube.com/watch?v=zwQmodmp1To)
- [video-use: drop raw footage in a folder, get final.mp4 back](https://thereelstack.com/lab/video-use-edit-footage-with-coding-agents): Browser Use's open-source editing skill reads your video as a transcript instead of watching it, then cuts, grades, captions and checks its own render. (spotted, not yet tested; source https://github.com/browser-use/video-use)
- [OpenMontage: an open-source agentic video production system](https://thereelstack.com/lab/openmontage-agentic-video-production): One of the most-starred agent video repos right now, describing itself as 12 production pipelines with 100+ tools. (spotted, not yet tested; source https://github.com/calesthio/OpenMontage)
- [Remotion + remotion-dev/skills: React video your agent can write](https://thereelstack.com/lab/remotion-and-remotion-skills): The most established code-to-video framework now ships official agent skills. Check the license before you build a business on it. (spotted, not yet tested; source https://www.remotion.dev/)
- [hypit: 'clone any viral video' with agents, and why we're cautious](https://thereelstack.com/lab/hypit-clone-viral-video-workflow): A fast-rising repo that rebuilds a viral video's whole workflow, including face swaps. Useful to study, risky to use carelessly. (spotted, not yet tested; source https://github.com/hypit-ai/hypit)
- [video-shotcraft: 152 cinematic shots as a Claude Code skill](https://thereelstack.com/lab/video-shotcraft-remotion-product-videos): A skill for Claude Code and Codex that builds cinematic product videos in Remotion from a library of 152 shot types. (spotted, not yet tested; source https://github.com/Vincentwei1021/video-shotcraft)
- [Sherpa to Seedance 2.5: how a 4.1M-view AI film was likely made](https://thereelstack.com/lab/sherpa-to-seedance-race-for-agi-reconstruction): The post credits only 'Script: Sherpa by Pocket FM, Video: Seedance 2.5'. This is our reconstruction of the probable pipeline, not the creator's confirmed process. (spotted, not yet tested; source https://x.com/maxxmalist/status/2102429352650514714)
- [The top GitHub video repos right now, ranked by stars (Oct 2026)](https://thereelstack.com/lab/top-github-video-repos-oct-2026): Twenty repos for AI video, agent video skills and ComfyUI workflows, pulled from the GitHub search API on Oct 14 2026. (spotted, not yet tested; source https://github.com/search?q=video+agent&type=repositories&s=stars)

## Directory
- [HyperFrames skills](https://github.com/heygen-com/hyperframes) [Skill, Code-to-video engine]: HTML → deterministic MP4; /hyperframes router picks launch, explainer, PR-to-video, recut or music workflows.
- [Remotion](https://github.com/remotion-dev/remotion) [Repo, Code-to-video engine]: React components → video. Mature, Lambda rendering; paid licence above 3 people.
- [Remotion agent skills](https://github.com/remotion-dev/skills) [Skill, Code-to-video engine]: Official Remotion best practices for Claude Code, Codex, Cursor.
- [Cetus](https://github.com/cenvero/cetus) [Repo, Code-to-video engine]: Single-binary HTML → MP4/WebM renderer with Chrome and ffmpeg bundled. Very early.
- [claude-motion](https://github.com/whaleyxbt/claude-motion) [Skill, Motion graphics]: Taste rules with numbers, contact-sheet self-review, Python sound engine, web sims → 60 fps.
- [animate](https://github.com/cth9191/animate) [Skill, Motion graphics]: Canvas + Playwright + FFmpeg animation in 7 styles; storyboard gates; Whisper word timing.
- [product-launch-motion](https://github.com/AbubakrChan/product-launch-motion) [Skill, Motion graphics]: Launch videos from code: GSAP, word-locked VO, camera moves, film grade, loudness.
- [motion-graphics-skills](https://github.com/charlie947/motion-graphics-skills) [Skill, Motion graphics]: 13 skills for launches, animated charts and reels.
- [claude-motion-design](https://github.com/howseen-ai/claude-motion-design) [Skill, Motion graphics]: HTML + Playwright + ffmpeg motion design; cuts on the beat.
- [Barty-Bart motion-graphics](https://github.com/Barty-Bart/motion-graphics) [Skill, Motion graphics]: Motion B-roll timed to your words over your own footage.
- [claude-remotion-skill](https://github.com/haidrrrry/claude-remotion-skill) [Skill, Motion graphics]: Remotion skill targeting flat fades and stiff motion.
- [gsap-skills](https://github.com/greensock/gsap-skills) [Skill, Motion graphics]: Official GSAP skills; the engine under many launch kits.
- [LottieFiles motion-design-skill](https://github.com/LottieFiles/motion-design-skill) [Skill, Motion graphics]: Timing, easing, choreography and Disney's 12 principles for agents.
- [motion-dev-animations-skill](https://github.com/199-biotechnologies/motion-dev-animations-skill) [Skill, Motion graphics]: Motion.dev web animation skill (springs, scroll, gestures).
- [Higgsfield MCP](https://higgsfield.ai/creator-hub/help-center/integrations/what-is-higgsfield-mcp) [MCP, Generation]: All Higgsfield models, Soul characters, audio, upscaling, Kling Motion Control inside Claude. Uses credits.
- [higgsfield-ai/skills](https://github.com/higgsfield-ai/skills) [Skill, Generation]: generate / soul / product-photoshoot CLI skills for coding agents.
- [Higgsfield Katana](https://higgsfield.ai/blog/higgsfield-katana) [Skill, VFX & compositing]: Rebuild a reference edit you own with up to 5 photos of your characters/products.
- [ElevenLabs MCP](https://github.com/elevenlabs/elevenlabs-mcp) [MCP, Audio & music]: Official MCP for TTS, voices and sound effects.
- [ComfyUI Wan 3.0 templates](https://docs.comfy.org/tutorials/partner-nodes/wan/wan3-0) [ComfyUI, Generation]: Official t2v / i2v / r2v graphs; @Image1/@Video1/@Audio1 reference tags (Partner Nodes, credits).
- [Comfy-Org workflow_templates](https://github.com/Comfy-Org/workflow_templates) [ComfyUI, Generation]: Official template repo (Wan 3.0, vCube enhance, Meshy 3D...).
- [ComfyUI-WanVideoWrapper](https://github.com/kijai/ComfyUI-WanVideoWrapper) [ComfyUI, Generation]: Community nodes for running Wan models locally.
- [ComfyUI-VideoHelperSuite](https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite) [ComfyUI, Generation]: Load/combine video nodes used in most ComfyUI video graphs.
- [LTX-Video](https://github.com/Lightricks/LTX-Video) [Repo, Generation]: Open video model family for self-hosting.
- [Wan 3.0 (official)](https://github.com/AlibabaCloud-Official/Wan3.0) [Repo, Generation]: 30 s native clips, up to 20 reference assets, instruction video editing.
- [ACE-Step 1.5](https://github.com/ace-step/ACE-Step-1.5) [Repo, Audio & music]: Local song generation, covers, stems, LoRA from a few songs, BPM/key extraction.
- [DiffRhythm](https://github.com/ASLP-lab/DiffRhythm) [Repo, Audio & music]: Lyrics → full-length song, fast.
- [MMAudio](https://github.com/hkchengrex/MMAudio) [Repo, Audio & music]: Video-to-audio synced foley (~6 GB VRAM). Check licence for commercial use.
- [HunyuanVideo-Foley](https://github.com/Tencent-Hunyuan/HunyuanVideo-Foley) [Repo, Audio & music]: Foley generation from video + text.
- [Demucs](https://github.com/facebookresearch/demucs) [Repo, Audio & music]: Stem separation, used for beat detection on drums and clean vocal timing.
- [LatentSync](https://github.com/bytedance/LatentSync) [Repo, Lip-sync & timing]: Open audio-conditioned lip-sync (v1.6, 512px). Consenting subjects only.
- [WhisperX](https://github.com/m-bain/whisperX) [Repo, Lip-sync & timing]: Word-level timestamps + diarization for captions and cut timing.
- [faster-whisper](https://github.com/SYSTRAN/faster-whisper) [Repo, Lip-sync & timing]: Fast local transcription; used by the animate skill for word timing.
- [Genjustsu Open Source Workflow](https://github.com/sirioberati/Genjustsu-Open-Source-Workflow) [Repo, VFX & compositing]: Depth + SAM 3 + Seedance subject replacement on your own footage, with a hashed mask-approval gate.
- [SlopMonster](https://github.com/ItsssssJack/SlopMonster) [Skill, Image & design]: Lint scripts for AI tells before voice-over.
- [PhotoCraft](https://github.com/storytold/photocraft) [Repo, Image & design]: Open-source Rust photo editor agents can drive, for fixing starting frames.
- [Compositor](https://github.com/robbietilton/Compositor) [Repo, Image & design]: Open-source Photoshop alternative for Mac.
- [Skillry](https://skillry.dev/) [Skill, Motion graphics]: 400+ in-house designed skills (video, slides, web, image); free + paid tiers.

## Tools: Video generation
- [Runway](https://thereelstack.com/tools/runway): preliminary score 72/100; official site https://runwayml.com
- [Google Veo](https://thereelstack.com/tools/google-veo): preliminary score 72/100; official site https://deepmind.google/models/veo/
- [Higgsfield](https://thereelstack.com/tools/higgsfield): preliminary score 70/100; official site https://higgsfield.ai
- [Kling](https://thereelstack.com/tools/kling): preliminary score 70/100; official site https://kling.ai
- [OpenAI Sora](https://thereelstack.com/tools/openai-sora): preliminary score 68/100; official site https://openai.com/sora
- [Luma Dream Machine](https://thereelstack.com/tools/luma-dream-machine): preliminary score 66/100; official site https://lumalabs.ai
- [Hailuo / MiniMax](https://thereelstack.com/tools/hailuo-minimax): preliminary score 64/100; official site https://hailuoai.video
- [Midjourney](https://thereelstack.com/tools/midjourney): preliminary score 64/100; official site https://www.midjourney.com
- [Adobe Firefly](https://thereelstack.com/tools/adobe-firefly): preliminary score 62/100; official site https://www.adobe.com/products/firefly.html
- [Pika](https://thereelstack.com/tools/pika): preliminary score 60/100; official site https://pika.art
- [Vidu](https://thereelstack.com/tools/vidu): preliminary score 58/100; official site https://www.vidu.com
- [LTX Studio](https://thereelstack.com/tools/ltx-studio): preliminary score 58/100; official site https://ltx.io/studio
- [Krea Video](https://thereelstack.com/tools/krea-video): preliminary score 58/100; official site https://www.krea.ai/video
- [Pollo AI](https://thereelstack.com/tools/pollo-ai): preliminary score 55/100; official site https://pollo.ai

## Tools: Avatars & talking heads
- [HeyGen](https://thereelstack.com/tools/heygen): preliminary score 70/100; official site https://www.heygen.com
- [Synthesia](https://thereelstack.com/tools/synthesia): preliminary score 68/100; official site https://www.synthesia.io
- [Hedra](https://thereelstack.com/tools/hedra): preliminary score 62/100; official site https://www.hedra.com
- [D-ID](https://thereelstack.com/tools/d-id): preliminary score 58/100; official site https://www.d-id.com
- [Colossyan](https://thereelstack.com/tools/colossyan): preliminary score 58/100; official site https://www.colossyan.com
- [Tavus](https://thereelstack.com/tools/tavus): preliminary score 58/100; official site https://www.tavus.io
- [Argil](https://thereelstack.com/tools/argil): preliminary score 55/100; official site https://www.argil.ai

## Tools: Editing & repurposing
- [DaVinci Resolve](https://thereelstack.com/tools/davinci-resolve): preliminary score 74/100; official site https://www.blackmagicdesign.com/products/davinciresolve
- [Descript](https://thereelstack.com/tools/descript): preliminary score 70/100; official site https://www.descript.com
- [CapCut](https://thereelstack.com/tools/capcut): preliminary score 66/100; official site https://www.capcut.com
- [OpusClip](https://thereelstack.com/tools/opusclip): preliminary score 64/100; official site https://www.opus.pro
- [VEED](https://thereelstack.com/tools/veed): preliminary score 62/100; official site https://www.veed.io
- [Kapwing](https://thereelstack.com/tools/kapwing): preliminary score 60/100; official site https://www.kapwing.com
- [InVideo](https://thereelstack.com/tools/invideo): preliminary score 56/100; official site https://invideo.io
- [Gling](https://thereelstack.com/tools/gling): preliminary score 56/100; official site https://www.gling.ai
- [Autopod](https://thereelstack.com/tools/autopod): preliminary score 54/100; official site https://www.autopod.fm
- [Wisecut](https://thereelstack.com/tools/wisecut): preliminary score 52/100; official site https://wisecut.ai
- [Cutback](https://thereelstack.com/tools/cutback): preliminary score 52/100; official site https://cutback.video

## Tools: Upscaling & enhancement
- [Topaz Video](https://thereelstack.com/tools/topaz-video): preliminary score 72/100; official site https://www.topazlabs.com/topaz-video
- [Magnific](https://thereelstack.com/tools/magnific): preliminary score 62/100; official site https://www.magnific.com

## Tools: Voice, music & dubbing
- [ElevenLabs](https://thereelstack.com/tools/elevenlabs): preliminary score 76/100; official site https://elevenlabs.io
- [Suno](https://thereelstack.com/tools/suno): preliminary score 66/100; official site https://suno.com
- [Udio](https://thereelstack.com/tools/udio): preliminary score 62/100; official site https://www.udio.com
- [Rask AI](https://thereelstack.com/tools/rask-ai): preliminary score 60/100; official site https://www.rask.ai
- [Vozo](https://thereelstack.com/tools/vozo): preliminary score 54/100; official site https://www.vozo.ai

## Tools: Captions
- [Captions](https://thereelstack.com/tools/captions): preliminary score 62/100; official site https://captions.ai
- [Submagic](https://thereelstack.com/tools/submagic): preliminary score 62/100; official site https://www.submagic.co

## Tools: Agent skills, MCPs & APIs
- [fal.ai](https://thereelstack.com/tools/fal-ai): preliminary score 66/100; official site https://fal.ai
- [Replicate](https://thereelstack.com/tools/replicate): preliminary score 66/100; official site https://replicate.com
- [ElevenLabs MCP](https://thereelstack.com/tools/elevenlabs-mcp): preliminary score 62/100; official site https://github.com/elevenlabs/elevenlabs-mcp
- [Smithery](https://thereelstack.com/tools/smithery): preliminary score 58/100; official site https://smithery.ai
- [WaveSpeed](https://thereelstack.com/tools/wavespeed): preliminary score 56/100; official site https://wavespeed.ai
- [Mootion](https://thereelstack.com/tools/mootion): preliminary score 52/100; official site https://www.mootion.com