Using Your Own AI Models (GGUF)
Hi everyone,
You can load your own AI models for experimentation, provided your system supports them.
Place your GGUF model file in:
CielChan → CielChan Desktop AI_Data → StreamingAssets
Launch CielChan, and your model will appear in the model dropdown.
Most GGUF models released before March 2026 should be compatible. Some newer models, including the latest Gemma releases, are not currently supported.
Multimodal models can be used for text interactions without additional setup. To enable their image-processing capabilities:
1. Select your custom model from the dropdown.
2. Open its model settings.
3. Click Select Vision Projector.
4. Choose the matching `mmproj` GGUF file supplied with the model.
5. Enable Vision.
The projector is saved separately for each custom model. To remove it, click Unload Vision Projector. If the projector is incompatible or fails to load, CielChan will remove the pairing and fall back to the built-in Lite Vision system when available.