comfyui-replicate

comfyui-replicate
★ 209

模型集成云端推理扩展节点快速安装
在 ComfyUI 中运行 Replicate 平台的模型,方便调用云端模型进行多样化推理与生成,扩展节点可用模型资源。
💡 在 ComfyUI 中调用 Replicate 模型进行云端推理与图像生成。
🍴 30 Forks💻 Python🔄 2024-11-05
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/8b0992d318c3
📦 requirements.txt
replicate>=1.0.3
example-screenshot
📄 README

comfyui-replicate

Custom nodes for running Replicate models in ComfyUI.

Take a look at the example workflows and the supported Replicate models to get started.

Set your Replicate API token before running

Make sure you set your REPLICATE_API_TOKEN in your environment. Get your API tokens here, we recommend creating a new one:

https://replicate.com/account/api-tokens

To pass in your API token when running ComfyUI you could do:

On MacOS or Linux:

export REPLICATE_API_TOKEN="r8_************"; python main.py

On Windows:

set REPLICATE_API_TOKEN="r8_************"; python main.py

Direct installation

cd ComfyUI/custom-nodes
git clone https://github.com/replicate/comfyui-replicate
cd comfyui-replicate
pip install -r requirements.txt

Supported Replicate models

View the supported_models.json to see which models are packaged by default.

Update Replicate models

Simply run ./import_schemas.py to update all model nodes. The latest version of a model is used by default.

Add more models

Only models that return simple text or image outputs are currently supported. If a model returns audio, video, JSON objects or a combination of outputs, the node will not work as expected.

If you want to add more models, you can:

  • add the model to supported_models.json (for example, fofr/consistent-character)
  • run ./import_schemas.py, this will update all schemas and import your new model
  • restart ComfyUI
  • use the model in workflow, it’ll have the title ‘Replicate [model author/model name]’
  • Roadmap

    Things to investigate and add to this custom node package:

  • support for more types of Replicate model (audio and video first)
  • showing logs, prediction status and progress (via tqdm)
  • Contributing

    If you add models that others would find useful, feel free to raise PRs.