A simple local image and video UI for ComfyUI, without the graph editor.
Quick start · AI install prompt · Update · Features · Website · ComfyUI · License
Fullscreen prompt-first generation.
Inspect an output and copy its settings.
Live ComfyUI previews while an image is running.
- Prompt-first image and video generation
- Model-aware controls from ComfyUI node metadata
- Image and video galleries kept separate by mode
- Zen mode for a cleaner fullscreen workflow
- Live queue/progress cards with cancel controls
- Persistent local gallery metadata
- Start-image reuse when the selected ComfyUI workflow supports it
- Importable ComfyUI API workflow templates
J AI Studio expects ComfyUI to already be installed and running.
npm install
npm run build
npm startOpen:
http://127.0.0.1:8787
By default, the app connects to ComfyUI at:
http://127.0.0.1:8188
Paste this into Codex, Claude Code, or another local coding agent:
Install and run J AI Studio from GitHub: https://github.com/jasperdevs/J-AI-Studio
Please do the full local setup for me:
1. Check whether Node.js 20+ is installed.
2. Check whether ComfyUI is installed and running at http://127.0.0.1:8188.
3. If ComfyUI is not running, help me start my existing ComfyUI install. Do not download models unless I explicitly ask.
4. Clone https://github.com/jasperdevs/J-AI-Studio into a normal projects folder.
5. Run npm install.
6. Copy .env.example to .env only if configuration changes are needed.
7. Set COMFY_URL to my ComfyUI URL, usually http://127.0.0.1:8188.
8. Run npm run build.
9. Start the app with npm start.
10. Open http://127.0.0.1:8787 and verify the app can reach ComfyUI, detect models, and load the gallery.
11. For future updates, use Settings -> Update, or run git pull, npm install, and npm run build.
Keep everything local. Do not expose HOST=0.0.0.0 unless I ask for phone or LAN access. If something fails, read the error, check ComfyUI /object_info and /system_stats, and fix the setup instead of guessing.
- Node.js 20 or newer
- A running ComfyUI server
- Local ComfyUI model files
From the app, open Settings -> Update to check GitHub and install the latest commit for a Git checkout. Restart the local server after an update.
CLI fallback:
git pull
npm install
npm run build
npm startTo check dependency updates:
npm run check:updatesJ AI Studio runs on top of ComfyUI. It reads available models, samplers, schedulers, size limits, prompt limits, text encoders, and VAEs from your local ComfyUI server where ComfyUI exposes them.
It does not replace ComfyUI, download models, train models, patch your ComfyUI install, or maintain a separate model runtime. ComfyUI remains the source of truth for installed nodes, model files, queue execution, previews, and output files.
Model support
The app is meant to be a simpler front end for common ComfyUI image and video generation, not a replacement for the graph editor.
Models appear when J AI Studio can detect enough ComfyUI metadata to build a generation workflow for them. If a model needs a custom graph, custom nodes, or special wiring, open it in ComfyUI first and confirm the required nodes are installed.
Workflow support is template-based. Built-in defaults cover common image/checkpoint/video paths. For anything else, export a ComfyUI API workflow and import it in Settings -> Workflows. Imported workflows appear only when their required ComfyUI nodes are installed.
For custom workflow files, see workflows/README.md.
Start-image controls only appear when the selected workflow exposes the required image input path through ComfyUI.
Generated files and model files stay local in your ComfyUI setup.
Configuration
Copy .env.example to .env if you need different ports or paths.
COMFY_URL=http://127.0.0.1:8188
HOST=127.0.0.1
PORT=8787
JAI_DATA_DIR=./data
COMFY_OUTPUT_DIR=COMFY_OUTPUT_DIR is optional. Set it only if you want the app's output-folder button to open a specific ComfyUI output directory.
Local network hosting
For another device on your network, set:
HOST=0.0.0.0Then open the selected PORT in your firewall. Only do this on a trusted network.
Windows shortcut example
You can make a shortcut that starts ComfyUI, starts J AI Studio, and opens the browser.
$appRoot = "C:\path\to\J-AI-Studio"
$comfyRoot = "C:\path\to\ComfyUI"
$python = "C:\path\to\python.exe"
if (-not (Get-NetTCPConnection -LocalPort 8188 -State Listen -ErrorAction SilentlyContinue)) {
Start-Process $python "main.py --listen 127.0.0.1 --port 8188 --disable-auto-launch" -WorkingDirectory $comfyRoot -WindowStyle Hidden
}
if (-not (Get-NetTCPConnection -LocalPort 8787 -State Listen -ErrorAction SilentlyContinue)) {
Start-Process node "server/index.js" -WorkingDirectory $appRoot -WindowStyle Hidden
}
Start-Process "http://127.0.0.1:8787/"npm install
npm run devThe dev command starts Vite and the local API server together.
See CONTRIBUTING.md. Do not commit generated media, local model files, logs, or .env files.
If no models appear, make sure ComfyUI is running and that COMFY_URL points to the right server.
If generation fails, confirm the selected model works in ComfyUI and that any required custom nodes are installed.
If video is missing, confirm your ComfyUI install has video generation nodes available.
MIT