diffusers-workflow
Active·★ 18·Apache-2.0·Updated 2026-10-20
Diffusers-workflow transforms the Hugging Face Diffusers library into a declarative engine, enabling image and video pipelines to be described as data. It provides an MCP server for agent interaction (e.g., Claude Code), a web UI, and a CLI/REPL for driving GPU-powered generative AI tasks.
#Diffusers#Generative AI#Image Generation#Video Generation#Workflow Automation#Hugging Face#Agent-driven#Web UI
01
Features
01MCP server integration: Enables AI agents (like Claude Code) to author, validate, run, and inspect generative tasks on a running server, with cost gating for GPU/disk operations.
02Web UI with introspection-driven forms: Provides a browser-based interface for managing workflows, editing them with forms generated from pipeline signatures, and monitoring live job progress.
03Reproducible and composable workflows: Each run is recorded with its full workflow definition, allowing easy reproduction, and supports multi-step pipelines for chaining image-to-video, inpainting, and ControlNet tasks.
04Comprehensive model and asset management: Includes a model manager for cached Hugging Face models, a shared prompt library, and asset management for promoting generated files.
05Cross-platform GPU acceleration: Supports various GPU backends including CUDA (NVIDIA), MPS (Apple Silicon), and CPU, along with quantization and inference acceleration techniques.
02
Compatibility
Python
Runtime
Verified via docs
NVIDIA CUDA
GPU Backend
Verified via docs
Apple Silicon MPS
GPU Backend
Verified via docs
CPU
Hardware
Verified via docs
03
Quick start
1
$ bash ./install.sh
2
$ source ./activate
04
Use cases
↳AI Agent-driven content generation: Allow agents like Claude Code to programmatically drive image and video generation on a GPU, managing complex workflows and monitoring progress.
↳Rapid prototyping and iteration of generative models: Use the Web UI or REPL to quickly test and iterate on different prompts, models, and pipeline configurations, reusing loaded models for faster iteration.
↳Building multi-stage generative AI pipelines: Compose complex workflows by chaining different Diffusers pipelines (e.g., text-to-image followed by image-to-video or inpainting).
↳Collaborative generative art/design projects: Share a single GPU server with multiple isolated workspaces, allowing different users or agents to work on their projects without interfering with each other.
05
Alternatives
comfyui-mcp★ 853
Claude Code plugin + MCP server for ComfyUI — 88 tools, 14 AI skills (Flux, WAN, LTX video, Qwen), live graph editing from your Claude session. Generate images & video, manage models and custom nodes.
blender-plugin★ 568
Scenario for Blender: AI images, video, 3D and PBR materials in the viewport, plus a local MCP server for agents. Blender 5.0+ extension, GPL-3.0-or-later.
blenderwright★ 175
MCP server for Blender. 191 tools for modelling, materials, rendering and more, driven by Claude, Cursor, Codex, Ollama, llama.cpp, or any MCP client.
gemini-mcp★ 34
Google Gemini MCP - Use Google's visual, video and chat capabilities in your AI Assistant. Uses MCP Apps to preview created images, landing pages, SVGs and video
superCMO-skills★ 138
Open-source skills that empower any AI agent (Claude, Cursor, Hermes, etc.) to generate end-to-end marketing campaigns - UGC videos, ad videos, product photography and more.
muapi-cli★ 12
Official CLI for muapi.ai — generate images, videos & audio from the terminal. MCP server, 14 AI models, npm + pip installable.
kreuzberg★ 9.4k
A polyglot document intelligence framework with a Rust core. Extract text, metadata, and structured information from PDFs, Office documents, images, and 50+ formats. Available for Rust, Python, Ruby, Java, Go, PHP, Elixir, C#, TypeScript (Node/Bun/Wasm/Deno)- or use via CLI, REST API, or MCP server.
Comments
Log in to leave a comment
No comments yet. Be the first!