ComfyStudio is a desktop animatic, previsualization, and AI-assisted editing tool built around a local ComfyUI workflow.
It combines a timeline editor, asset browser, Generate workspace, stock search, LM Studio prompt helper, and export queue in one app.
Most users should download the desktop app from the GitHub Releases page.
The current public release is marked as a pre-release while early setup feedback comes in.
Desktop app release assets include:
- Windows installer
- Windows portable build
- macOS Apple Silicon DMG
- macOS Intel DMG
Advanced ComfyUI users can also optionally download the Workflow Starter Pack release asset to inspect workflows directly in ComfyUI and prepare dependencies manually before queueing inside ComfyStudio.
- The ComfyStudio desktop app
- Built-in workflow JSONs used by the app internally
- In-app dependency checks, setup guidance, and the Getting Started guide
- ComfyUI itself
- Every custom node and model required for every workflow
- Real credentials such as Comfy account API keys or Pexels keys
- Edit projects with assets, preview, timeline, inspector, and export tools.
- Generate images, videos, and audio through curated local or cloud workflows.
- Use Director Mode to turn structured script input into keyframes and video shots.
- Search Pexels stock footage and stills directly from the app.
- Refine prompts locally with LM Studio.
- ComfyUI is local-only in this build. Remote and LAN ComfyUI targets are intentionally disabled.
- The default ComfyUI endpoint is
http://127.0.0.1:8188, but users can override the port in Settings. - LM Studio integration is local-only.
- Generate includes dependency preflight checks before queueing workflows.
Minimum for normal app use:
- ComfyUI installed separately and running locally
Optional integrations:
- Pexels API key for the
Stocktab - Comfy account API key for paid partner-node workflows
- LM Studio for local prompt assistance in the
LLMtab
When you first open the app:
- Choose a projects folder.
- Create or open a project.
- Open
ComfyStudio > Getting Startedfrom the bottom menu for the built-in setup and orientation guide.
The Getting Started guide is manual by design. It does not auto-open.
ComfyStudio ships with built-in workflow JSON files, but that does not mean every workflow will work immediately on every machine.
Users still need the correct combination of:
- ComfyUI custom nodes
- model files
- local GPU capacity for local workflows
- partner API credentials for cloud workflows
- Default:
8188 - Open
Settings > ComfyUI Connectionto change or test the local port. - Only localhost or loopback addresses are accepted in this build.
Inside Generate, choose a workflow and use the dependency check tools if something is missing.
The current workflow help flow includes:
- dependency preflight before queueing
Re-checkCopy reportOpen node registryOpen in ComfyUI
Use this for one-off image, video, or audio generations.
Use this for a guided sequence:
SetupScriptKeyframesVideos
Director Mode supports a structured script format so the app can parse scene context, shot type, keyframe prompt, motion prompt, camera direction, and duration.
Example:
Scene 1: Neon Arrival
Scene context: Futuristic transit terminal, blue and coral neon, reflective black tile, premium cinematic sneaker ad.
Shot 1:
Shot type: Wide shot
Keyframe prompt: Wide shot of the model stepping through sliding glass doors into a futuristic transit terminal, blue and coral neon reflecting across glossy black tile, coral-and-cream sneaker clearly visible.
Motion prompt: Starting from this exact keyframe, the model takes 2 confident steps forward while neon reflections slide across the floor. Keep the sneaker, outfit, and terminal lighting consistent.
Camera: Gentle backward tracking shot
Duration: 3
ComfyStudio tags workflows in Generate with a hardware tier to make setup expectations clearer.
| Tier | Meaning | Typical workflows | Practical guidance |
|---|---|---|---|
| Lite | Lower-end local GPU | Z Image Turbo, Music Generation | Often workable on 6-8 GB VRAM |
| Standard | Mid-range local GPU | Image Edit, Multiple Angles | Often needs 12-16 GB VRAM |
| Pro | Higher-end local GPU | WAN 2.2 image-to-video | Usually wants 24 GB+ VRAM |
| Cloud | Partner-node / credit-backed | Nano Banana 2 and other cloud workflows | Local VRAM is not the main constraint |
VRAM guidance is approximate. Resolution, frame count, model variant, batch size, and other running apps can change the real requirement significantly.
Cloud workflows can display:
- estimated credits per run
- approximate USD conversion
- projected plan totals in Director Mode
For some workflows, pricing is dynamic. In those cases the UI shows a dynamic pricing label when the estimate is not fixed. Live account balance is not currently shown in-app.
The Stock tab uses Pexels.
- Add a Pexels API key in
Settings. - Search and import photos or videos directly into the current project.
The LLM tab connects to LM Studio.
- Start LM Studio locally.
- Enable the local API server.
- Load a model.
- Use the model to refine prompts before generating.
If your GPU memory is tight, unload the LM Studio model before heavy ComfyUI generation.
The Workflow Starter Pack is optional and intended for advanced ComfyUI users who want a clearer dependency and workflow-prep path outside the app.
Advanced ComfyUI users can usually treat it like this:
- Download the Workflow Starter Pack
- Open the workflows in ComfyUI
- Install any missing custom nodes
- Download any missing models
- Return to ComfyStudio and use
Generate
Generate it with:
npm run starter-pack:build
npm run starter-pack:packageGenerated files:
docs/workflow-starter-pack/starter-pack.manifest.jsondocs/workflow-starter-pack/INDEX.mddocs/workflow-starter-pack/docs/workflows/*.mddocs/workflow-starter-pack/workflows/local/*.comfyui.jsondocs/workflow-starter-pack/workflows/cloud/*.comfyui.json
See:
docs/workflow-starter-pack/README.mddocs/workflow-starter-pack/INDEX.mddocs/workflow-starter-pack/docs/WHERE_FILES_GO.mddocs/workflow-starter-pack/docs/API_KEYS.mddocs/workflow-starter-pack/docs/TROUBLESHOOTING.md
If you want to develop or contribute to ComfyStudio, run it from source with Electron:
npm install
npm run electron:devnpm install
npm run electron:devThe packaged desktop app is the intended experience for most users. Browser-only npm run dev is still useful for frontend work, but Electron is the normal development path.
npm run build
npm run electron:build:win
npm run electron:build:mac
npm run electron:build:linuxPackaged artifacts are written to release/.
- ComfyUI is local-only in this build.
- Some workflows require manual model or custom-node installation on the user's ComfyUI machine.
- Cloud pricing can be dynamic for certain partner workflows.
- Live partner credit balance is not currently shown in-app.
See CONTRIBUTING.md.
MIT. See LICENSE.