NOTE: Since native NDI support got integrated in Daydream Scope, this bridge has no utility anymore. Still useful to use NDI with StreamDiffusion via the Daydream API

Daydream StreamDiffusionV1 NDI Bridge

Real-time AI video transformation bridge using the Daydream API. Capture video via NDI and transform it with AI in real-time.

resolume_github.mp4

Features

• NDI Input - Capture video from any NDI source on your network
• Real-time AI Transformation - Transform your video feed using StreamDiffusion
• Multiple AI Models - Support for SD 1.5, SD Turbo, SDXL Turbo
• ControlNet Support - Depth, Canny edge, and Tile ControlNets
• Web-based Control Panel - Adjust all parameters in a clean browser UI
• Low Latency - WebRTC output for minimal delay

Requirements

• macOS 10.15+ or Windows 10+ (with NDI SDK)
• Python 3.10+
• NDI Tools installed
• Active internet connection

Installation

cd daydream-bridge

# Install dependencies
pip install -r requirements.txt

# Run the bridge
python app.py

Authentication

On first run, a browser window will open to sign in to Daydream:

1. Create an account at daydream.live if you don't have one. The free trial is 10 hours of video.
2. Sign in when the browser opens
3. The app will automatically create and save an API key to ~/.daydream/credentials

You only need to sign in once — the API key is saved locally and reused on subsequent runs.

Usage

1. Start your NDI source (VirtualDJ, Resolume, OBS, or any NDI-capable app)
2. Run the bridge with python app.py
3. Open the control panel in your browser (opens automatically)
4. Select your NDI source from the dropdown
5. Click "Start Stream" to begin AI transformation
6. Adjust parameters in real-time using the sliders

Controls

Parameter	Description
Prompt	Text description of desired visual style
Negative Prompt	What to avoid in generation
Denoise (Delta)	How much AI transformation to apply
Depth	Preserve depth/3D structure
Canny	Preserve edges/outlines
Tile	Preserve texture patterns
Guidance	How closely to follow the prompt

NDI Sources

The bridge automatically scans for NDI sources on your network. Common sources include:

• VirtualDJ - Enable NDI output in VirtualDJ's broadcast settings
• OBS Studio - Use the NDI plugin
• Any NDI-capable application

Backend Options

The bridge supports two backends:

1. Daydream Cloud (Default)

Uses the Daydream cloud API for processing. Easy to set up, just sign in and go.

• 10 hours free trial
• No GPU required
• StreamDiffusion v1 models

2. Scope (Self-Hosted or RunPod)

Use Daydream Scope for more models and no cloud costs.

Features:

• StreamDiffusionV2, LongLive, Krea Realtime Video, and more
• LoRA support for custom styles
• Run locally or on RunPod

Using Scope with this bridge:

1. Start Scope (locally or on RunPod)

   • Local: uv run daydream-scope → http://localhost:8000
   • RunPod: Use the RunPod template → get your proxy URL

2. In the bridge control panel:

   • Switch to "Scope (Self-hosted)" tab
   • Paste your Scope URL (e.g., https://xxx-8000.proxy.runpod.net)
   • Click Test Connection to verify
   • Select NDI source and start streaming

RunPod setup:

1. Deploy the Scope RunPod template
2. Set HF_TOKEN environment variable for TURN servers (required for WebRTC through firewalls)
3. Wait for deployment, get your proxy URL from RunPod
4. Paste the URL in the bridge and stream!

GPU requirements for self-hosting:

• Minimum: RTX 3090/4090 (24GB VRAM)
• For Krea Realtime Video: RTX 5090 or better (32GB+ VRAM)

License

MIT License - see LICENSE file for details.

Credits

• Daydream - Real-time AI video generation API
• Daydream Scope - Self-hosted AI video generation
• NDI - Network Device Interface