Hosted StreamDiffusion for TouchDesigner

StreamDiffusionTD is a TouchDesigner operator that connects real time inputs like camera feeds, audio, and sensors to StreamDiffusion, turning them into live generative visuals inside TouchDesigner. It was built by Lyell Hintz (known as dotsimulate) and is distributed through Dotsimulate's Patreon.

Hosted StreamDiffusion runs the StreamDiffusion pipeline on a remote GPU instead of your local machine. With the Daydream API, StreamDiffusionTD offloads inference to Daydream's cloud backend, so you get the same real time generation without a local NVIDIA GPU, Python environment, or CUDA install.

Only if you run it locally. The local path needs a CUDA capable NVIDIA GPU on Windows, with 6 GB of VRAM as the floor and 24 GB recommended for the full feature set. Professional quality output has generally meant an RTX 4090. Running StreamDiffusionTD against Daydream's hosted backend removes that requirement. Any machine that can run TouchDesigner and reach the internet will work, including Apple Silicon Macs.

StreamDiffusionTD runs SDXL Turbo as its native model for real time generation. On top of that, the hosted flow supports multi ControlNet, IP Adapter (including FaceID), and TensorRT acceleration, with real time control over prompt, seed, and ControlNet weights while your stream is live.

Yes. The Daydream hosted path supports multi ControlNet and IP Adapter with FaceID out of the box, and every parameter you would normally tune locally stays adjustable in real time through the plugin.

You need three things. First, the StreamDiffusionTD plugin from Dotsimulate's Patreon. Second, a free Daydream API key from the Builder Dashboard. Third, TouchDesigner installed (2023 or newer on Mac, 2025 or newer on Windows for the official Daydream quickstart file). The fastest path is to download Andrew Sun's Daydream SDXL Easy Quickstart File, which wires StreamDiffusionTD to your Daydream API key with a click and drag install. Total setup time is under a minute.

You can start with a free Daydream API key from the Builder Dashboard. Full pricing details, including paid tiers, live at daydream.live/pricing. The StreamDiffusionTD plugin itself is distributed through Dotsimulate's Patreon.

Locally, StreamDiffusionTD needs Windows, Python 3.10 or 3.11, a matching CUDA toolchain, the NDI SDK, and an NVIDIA GPU with enough VRAM for your model choice. Hosted swaps all of that for a single API key. The tradeoffs are straightforward. Local gives you zero network inference and full offline control. Hosted gives you cross platform support (macOS included), no GPU spend, and a one minute first run.