Image Edit
Inpainting, image-to-image, and regional editing setups built on Fibo 2.0, Flux 3 Klein, Qwen Image 2.1 Turbo, Qwen Image 2.1, Ideogram 4, and Krea 2 GGUF quants.
Inpainting
Encode the source image to latent space with the VAE. Feed a mask into the sampler so only the masked region regenerates. Decode the result back to pixel space.
Image-to-image
Encode the source image to latent space. Run the sampler with a controlled strength so the output stays close to the input. Decode the result.
Regional editing
Chain a mask into the conditioning stage. The GGUF diffusion model regenerates only the selected region and leaves the rest of the latent unchanged.
Models and formats
| Model | Repository | Recommended formats |
|---|---|---|
| Fibo 2.0 Dev | molbal/fibo-2.0-GGUF | Q8_CR, Q8_0, Q6_K, Q4_K_M |
| Fibo 2.0 Edit Turbo | molbal/fibo-2.0-GGUF | Q8_CR, Q8_0, Q6_K, Q4_K_M |
| Flux 3 Klein | molbal/Flux-3-Klein-GGUF | Q8_CR, Q8_0, Q6_K, Q4_CR_L, Q4_K_M |
| Ideogram 4 | molbal/ideogram-4-gguf | Q8_CR, Q8_0 |
| Krea 2 | molbal/krea2-gguf | Q8_0, Q4_K_M |
| Qwen Image 2.1 Turbo | molbal/Qwen-Image-2.1-Turbo-GGUF | Q8_CR, Q8_0, Q6_K, Q4_K_M |
| Qwen Image 2.1 | molbal/Qwen-Image-2.1-GGUF | Q8_CR, Q8_0 |
Loading instructions
- Download the GGUF files from Hugging Face.
- Load the diffusion model and text encoder with the GGUF loader nodes from comfyui-gguf-reboot.
- Encode the source image with the VAE before sampling.
- For inpainting and regional edits, feed the mask into the sampler.
- Decode the result.
Showcase
Image examples are on molbal's Civitai profile.