# chengzeyi/Comfy-WaveSpeed

https://wavespeed.ai/ [WIP] The all in one inference optimization solution for ComfyUI, universal, flexible, and fast.

Repository: https://github.com/chengzeyi/Comfy-WaveSpeed
Canonical: https://ross.abutalabs.com/products/comfy-wavespeed
Language: Python
License: MIT
License Family: permissive
Last push: 2026-08-21T05:35:11+00:00

## Health v2 (maintenance only)
Score: 66/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 98, release rhythm 35, longevity 46
- inputs: {"age_days": 644, "days_push": 12, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1231, forks 68 (observed 2026-08-28T04:04:04.222002+00:00)

## What it is
A ComfyUI custom node plugin that acts as an all-in-one inference optimization solution for diffusion models, built around First Block Cache (dynamic caching) and an enhanced torch.compile. It accelerates image and video generation models such as FLUX.1, Stable Diffusion 3.5, SDXL, HunyuanVideo, and LTXV directly inside ComfyUI workflows.

## Use cases
- speed up image generation in ComfyUI
- make FLUX render faster in ComfyUI
- reduce Stable Diffusion inference time
- apply first block cache to ComfyUI workflows
- accelerate diffusion model inference with torch.compile
- speed up HunyuanVideo or LTXV video generation
- get faster ComfyUI renders without changing the output

## When to choose
- You run ComfyUI and want significantly faster diffusion model inference with minimal or no quality loss
- You generate images or videos with FLUX, SD3.5, SDXL, HunyuanVideo, or LTXV and want cache- and compilation-based speedups
- Your GPU and software stack meet the torch.compile requirements and you want the maximum possible speedup

## When to avoid
- You do not use ComfyUI — this is a ComfyUI custom node, not a standalone inference optimizer
- You need multi-GPU or context-parallel inference, which is considered but not implemented
- You require a rapidly evolving feature set, since active development on this repository is intermittent

## Facets
- artifact type: plugin
- maturity: stable
- function: caching, stable-diffusion, deep-learning, gpu-computing
- domain: artificial-intelligence, deep-learning, image-processing
- platform: cross-platform, windows, python
- tags: comfyui, comfyui-custom-node, inference-optimization, first-block-cache, fbcache, dynamic-caching, torch-compile, diffusion-models, flux, sdxl, hunyuan-video, image-generation, video-generation, para-attention, speedup, video, linux, macos, desktop, gpu

## Member repositories
- chengzeyi/Comfy-WaveSpeed (main) score 66

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:04.222002+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T08:22:29.267309+00:00, confidence not recorded.
  - readme: https://github.com/chengzeyi/Comfy-WaveSpeed (fetched 2026-08-28T04:04:04.222002+00:00, sha f172c20a6d74)
- Data as of 2026-08-30T08:39:29.467469+00:00.