Unplatform AI Provider

<h1> <img src="docs/images/header-light.svg#gh-light-mode-only" height="56" alt="Unplatform AI Provider"> <img src="docs/images/header-dark.svg#gh-dark-mode-only" height="56" alt="Unplatform AI Provider"> </h1>

Turn any ComfyUI workflow into a model for Unplatform AI. Save it, add one URL in the app, generate.

<img src="docs/images/hero-dark.svg" alt="Unplatform AI connects through the UPAI node pack to a local ComfyUI endpoint, shown in the app's model picker." width="100%">

Quick start

1. Install the pack from ComfyUI Manager, search "Unplatform AI Provider". See getting started.
2. Open a workflow from workflows/ and save it in ComfyUI.
3. Add the ComfyUI URL in Unplatform AI. See getting started.

Workflows

<table> <thead> <tr> <th width="90"></th> <th align="left">Model</th> <th align="left">Kind</th> <th align="left">Modes</th> <th align="left">VRAM (full speed)</th> <th align="left">Docs</th> </tr> </thead> <tbody> <tr> <td width="90" align="center"><a href="workflows/flux-2-klein-readme.md"><img src="docs/images/flux-2-klein.png" width="64" alt=""></a></td> <td>FLUX.2 Klein</td> <td>Image</td> <td>Text-to-image / image-to-image with up to 4 references.</td> <td>16 GB</td> <td><a href="workflows/flux-2-klein-upai.json">JSON</a><br><a href="workflows/flux-2-klein-readme.md">Guide</a><br><a href="workflows/flux-2-klein-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/ideogram-4-readme.md"><img src="docs/images/ideogram-4.png" width="64" alt=""></a></td> <td>Ideogram 4</td> <td>Image</td> <td>Text-to-image.</td> <td>24 GB</td> <td><a href="workflows/ideogram-4-upai.json">JSON</a><br><a href="workflows/ideogram-4-readme.md">Guide</a><br><a href="workflows/ideogram-4-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/krea-2-readme.md"><img src="docs/images/krea-2.png" width="64" alt=""></a></td> <td>Krea 2</td> <td>Image</td> <td>Text-to-image / image-to-image with up to 2 references.</td> <td>24 GB</td> <td><a href="workflows/krea-2-upai.json">JSON</a><br><a href="workflows/krea-2-readme.md">Guide</a><br><a href="workflows/krea-2-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/ltx-2-5-readme.md"><img src="docs/images/ltx-2-5.png" width="64" alt=""></a></td> <td>LTX 2.5<br><small>LTX 2.5 Extend<br>LTX 2.5 Upscale</small></td> <td>Video</td> <td>Text-to-video / image-to-video / first-last-frame video with up to 2 frame references. Generates audio.<br>Video-to-video extend by up to 10 seconds with 1 video reference. Audio continues.<br>Video-to-video upscale to 720p or 1080p with 1 video reference. Audio is kept.</td> <td>32 GB</td> <td><a href="workflows/ltx-2-5-upai.json">JSON</a><br><a href="workflows/ltx-2-5-readme.md">Guide</a><br><a href="workflows/ltx-2-5-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/minimax-h3-readme.md"><img src="docs/images/minimax-h3.png" width="64" alt=""></a></td> <td>MiniMax H3<br><small>MiniMax H3 Fast<br>MiniMax H3 Extend<br>MiniMax H3 Image<br>MiniMax H3 Image Fast</small></td> <td>Video, Image</td> <td>Text-to-video / image-to-video with up to 9 image, 3 video and 3 audio references.<br>Fast text-to-video / image-to-video with up to 9 image, 3 video and 3 audio references.<br>Video-to-video extend by up to 15 seconds with 1 video reference. Audio continues.<br>Text-to-image / image-to-image with up to 16 references.<br>Fast 2-step text-to-image / image-to-image with up to 16 references.</td> <td>32 GB</td> <td><a href="workflows/minimax-h3-upai.json">JSON</a><br><a href="workflows/minimax-h3-readme.md">Guide</a><br><a href="workflows/minimax-h3-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/moss-image-readme.md"><img src="docs/images/moss-image.png" width="64" alt=""></a></td> <td>MOSS Image<br><small>MOSS Pose</small></td> <td>Image</td> <td>Text-to-image / image-to-image.<br>Image-to-image with 1 pose reference.</td> <td>12 GB</td> <td><a href="workflows/moss-image-upai.json">JSON</a><br><a href="workflows/moss-image-readme.md">Guide</a><br><a href="workflows/moss-image-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/qwen-2-1-readme.md"><img src="docs/images/qwen-2-1.png" width="64" alt=""></a></td> <td>Qwen 2.1<br><small>Qwen 2.1 Head Swap<br>Qwen 2.1 Body Swap<br>Qwen 2.1 Outpaint</small></td> <td>Image</td> <td>Text-to-image / image-to-image with up to 10 references.<br>Image-to-image with 2 references. No prompt.<br>Image-to-image with 2 references. No prompt.<br>Image-to-image outpaint to the selected resolution and aspect. No prompt.</td> <td>12 GB</td> <td><a href="workflows/qwen-2-1-upai.json">JSON</a><br><a href="workflows/qwen-2-1-readme.md">Guide</a><br><a href="workflows/qwen-2-1-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/wan-2-2-readme.md"><img src="docs/images/wan-2-2.png" width="64" alt=""></a></td> <td>Wan 2.2</td> <td>Video</td> <td>Text-to-video / image-to-video / first-last-frame video. Silent.</td> <td>24 GB</td> <td><a href="workflows/wan-2-2-upai.json">JSON</a><br><a href="workflows/wan-2-2-readme.md">Guide</a><br><a href="workflows/wan-2-2-skill.md">Skill</a></td> </tr> <tr> <td width="90" align="center"><a href="workflows/z-image-readme.md"><img src="docs/images/z-image.png" width="64" alt=""></a></td> <td>Z-Image</td> <td>Image</td> <td>Text-to-image / image-to-image.</td> <td>12 GB</td> <td><a href="workflows/z-image-upai.json">JSON</a><br><a href="workflows/z-image-readme.md">Guide</a><br><a href="workflows/z-image-skill.md">Skill</a></td> </tr> </tbody> </table> <p>MiniMax H3 Image license excludes use in the US, EU, UK and South Korea. Check the license before use.</p>

Smaller cards run with offload to system RAM, slower. RAM must hold the models.

9 workflows, 19 models (12 image, 7 video). Some workflows hold several models.

Docs

• Getting started
• Node reference
• Build a workflow
• Network
• Memory
• Troubleshooting
• FAQ

For developers

• Architecture
• Development
• Protocol spec

License

GPL-3.0 for the pack. protocol/ is MIT.

Not affiliated with ComfyUI or any model provider.