> ## Documentation Index
> Fetch the complete documentation index at: https://docs.protoface.com/llms.txt
> Use this file to discover all available pages before exploring further.

# ComfyUI

> Generate images, video, and audio from your ComfyUI workflow with Protoface.

The official [Protoface extension](https://protoface.com/comfyui) adds model-specific nodes to ComfyUI.
Generation runs through the Protoface API, so you do not need downloaded
model weights or a local GPU for Protoface nodes. Connect prompts and
reference media, then preview, save, or pass the results to another node.

You need **ComfyUI 0.41.0 or newer**, Python **3.10 or newer**, and an API
key from the [dashboard](https://app.protoface.com). Generations use your
Protoface API credits at the same [model rates](https://protoface.com/pricing).
The extension is free to install.

## Install with ComfyUI Manager

1. Open **Extensions** or **Manager** in ComfyUI.
2. Search **Protoface** and install **protoface-nodes**, published by
   **protoface** on [Comfy Registry](https://registry.comfy.org/nodes/protoface-nodes).
3. Restart ComfyUI and reload the browser.
4. Click **Protoface** in the lower-right corner and enter your API key.

If Manager does not show the package, refresh its catalog. For Git, ZIP,
Windows portable, and Desktop installation instructions, use the
[installation guide](https://github.com/protoface-ai/protoface-comfyui/blob/master/docs/installation.md).

The key stays in the ComfyUI server's memory and is excluded from exported
workflows. Enter it again after restarting the server. Anyone who can run
workflows on a shared server can use its configured key.

## Make your first generation

Open **Templates → the Protoface extension** and choose an included workflow,
or search **Protoface**, a model name such as **MiniMax H3** or **LTX 2.5**,
or a task such as **Text to Video** in Add Node.

Enter your prompt, set the model controls, and queue the graph. Native
image, video, and audio connections let you use reference media and feed
results into the next step. **Protoface Reuse Result Asset** passes an
existing result to another Protoface generation without reuploading it.

The [workflow guide](https://github.com/protoface-ai/protoface-comfyui/tree/master/example_workflows)
explains the included video, reference-media, asset-reuse, and recovery
templates. Model nodes become searchable after you install the extension;
Registry and Manager search find the package itself.

## Request a new generation or recover a run

Keep **generation\_id** fixed to reuse or resume the same request. Increment
it to request a fresh paid generation. Changing the prompt, model controls,
or reference inputs also creates a new request. The generation ID is
separate from the model's random seed.

If execution is interrupted, preserve the workflow and generation ID, then
requeue it. **Protoface Resume Run** retrieves an existing run by its run ID
without submitting another generation. Preserve the extension's local state
directory across updates so it can recover recorded jobs. Cancellation is
best effort and does not promise a refund.

## Update models and the extension

For compatible new models or controls, click **Protoface → Refresh catalog**,
then restart ComfyUI and reload the browser. These updates do not require
reinstalling the extension. Protoface settings lists any skipped tasks that
need a new adapter; compatible tasks still load.

For extension updates, use Manager to install the latest approved Registry
version, then restart and reload. A new control type or media adapter may
require an extension update. Keep only one installed copy.

## Headless and hosted use

For a persistent or headless setup, set `PROTOFACE_API_KEY_FILE` to a private
file containing your key before launching ComfyUI. You can also set
`PROTOFACE_API_KEY` in the server environment. Keep credentials outside
workflow JSON. API-format workflows need a stable, unique `request_scope`
for each workflow/node; preserve it when recovering a request.

Local ComfyUI and hosted services that permit independent custom nodes can
use this extension. Desktop users should follow the managed-environment
instructions in the installation guide. **Comfy Cloud does not currently
permit installing the Protoface extension**, and Comfy credits do not pay
for Protoface API generations.

## Troubleshooting

| Problem | What to check |
| - | - |
| No Protoface nodes | Restart and reload, check the startup log, and confirm there is only one copy installed. |
| Authentication or credit error | Re-enter your API key after a restart and check your account's API credits. |
| A new model is missing | Refresh the catalog, restart, and reload. Check Protoface settings for skipped tasks. |
| A run times out | Keep the generation ID fixed and requeue, or use Resume Run with its run ID. |

See the [full troubleshooting guide](https://github.com/protoface-ai/protoface-comfyui/blob/master/docs/installation.md#troubleshooting).
For support, include the extension version and run ID, but never your API key.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.