Comfy releases Comfy Router to manage image and video generation AI APIs in one place

Comfy Router operates on Comfy's developer platform and provides access to multiple state-of-the-art models through a single API. A key feature is the ability to switch between providers when the same model is offered by multiple sources. Even if a provider is busy or hits rate limits (restrictions on the number of requests allowed within a certain period), users can easily switch to a different provider by simply changing part of the API key to continue generation.

Comfy lists the following three primary use cases:

• Integration into Product Features: If access surges for a specific service and reaches rate limits, you can continue processing by switching to another provider simply by changing a single parameter.
• Workflow Pipelining: Workflows created in ComfyUI can be combined and executed with video generation models on Comfy Router using the same code and API key.
• Large-scale Batch Processing: Using the "submit" function, you can add hundreds of jobs to a queue and retrieve results as they are completed.

Initial supported models include ByteDance's "Seedance 2.5", MiniMax's "MiniMax H3", Google's "Nano Banana Pro", OpenAI's "ChatGPT Images 2.0", Black Forest Labs' "Flux", and Kuaishou's "Kling". Additional models are planned for future release.