Tongflow

TongFlow is a solid option if you're a developer or researcher who wants to own your multi-modal AI pipeline from end to end. The infinite canvas visual editor is a nice abstraction for composing complex transforms, and the local-first approach means your data never leaves your machine. We'd reach for this when building custom generative projects with multiple modalities—text, image, audio, video—and you want to experiment with the latest open models without cloud markups. But it's not a plug-and-play product. You'll need comfort with command-line tools, Node.js, Modal for GPU compute, and managing API keys for LLM providers. There's no cloud-hosted version, no mobile app, and no official support. The tool's audience is narrow: open-source developers and AI tinkerers. Compared to ComfyUI, TongFlow is more opinionated about local-first storage and multi-modal workflow; ComfyUI is more mature and has a larger node ecosystem. ComfyUI also has better documentation and community. But TongFlow pulls ahead with built-in support for audio and video models, which ComfyUI handles less gracefully. Where it bites: the setup friction. You need Modal plus at least one LLM API key just to get started—there's no free tier that works out of the box. The GitHub repo has minimal docs, so you'll be reading source code or reverse-engineering examples. That's fine for the target audience, but it's a hard pass for anyone wanting a ready-to-use studio. In practice, TongFlow is best for prototyping multi-modal pipelines locally, especially if you value data sovereignty. The AGPL-3.0 license is restrictive for commercial embedding; if that's a concern, look elsewhere.