Fibo 2.0

Fibo 2.0

Edits images in ComfyUI with repackaged Fibo 2.0 files from BRIA AI, in bf16 and int8 versions with Dev and Turbo edit workflows. For ComfyUI users working on their own machine. Non-commercial license only.

Made by Comfy-Org on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: This is a set of repackaged files for running BRIA AI's Fibo 2.0 image-editing models in ComfyUI. They cover the Dev and Edit Turbo versions, each as a bf16 file and a smaller int8 file. The repository also holds a Qwen3-VL 4B text encoder and a Flux 2 VAE to go with them.

Who it suits: ComfyUI users who want to edit images with Fibo 2.0 on their own computer, for non-commercial work. Workflow builders can also use the files as a base for their own graphs.

Industry | Who uses it | How they use it | Example
Image editing | ComfyUI users | Run Fibo 2.0 edits inside a node workflow | Open the Turbo edit workflow and change a picture
Creative hobby | Non-commercial creators | Edit images on their own machine | Choose the int8 files when you want a lighter copy
Research | ML researchers | Compare the bf16 files with the int8 files | Run both versions on one edit and compare the results
Tooling | Workflow builders | Build ComfyUI graphs on the repackaged files | Swap the Dev model for the Turbo one in a workflow

You can skip it if you do not use ComfyUI, or if you need commercial use without a separate agreement with BRIA AI.

Where to use it:

• ComfyUI, the node-based image tool, on Windows, macOS or Linux
• Your own computer, using files fetched with the Hugging Face tool
• The Dev and Turbo image edit workflow templates made for ComfyUI

Use case | Who | What they get back
Edit with the Dev model | ComfyUI users | An edited version of the picture you feed in
Edit with the Turbo model | ComfyUI users | An edited picture from the Turbo edit model
Version comparison | Researchers | Slight differences between bf16 and int8 results

How to set it up: You need ComfyUI and Python for the download tool.

• ComfyUI installed on your computer
• Python, to run the Hugging Face tool
• Disk space for the files you pick, since the card gives no sizes
• A non-commercial use, which the license requires

1. Install ComfyUI if you do not have it.
2. Install the Hugging Face hub tool and download the repository.
3. Move the diffusion model files into the diffusion models folder inside the ComfyUI models folder.
4. Move the text encoder files into the text encoders folder and the VAE file into the vae folder.
5. Pick the bf16 or int8 version of each part.
6. Open one of the two edit workflow templates in ComfyUI.

pip install -U huggingface_hub
hf download Comfy-Org/Fibo-2.0

How to use it: With the files in place, open the Dev or Turbo image edit workflow in ComfyUI and run it. The card shows a before and after picture for each workflow. It gives no commands beyond the download, so the workflow's own nodes show what to set.

Tool | What it does
Fibo 2.0 Dev: Image Edit | A ComfyUI workflow that edits an image with the Dev model
Fibo 2.0 Turbo: Image Edit | A ComfyUI workflow that edits an image with the Turbo edit model

Example: You put the Turbo files in their folders and open the Turbo image edit workflow in ComfyUI. You give it a picture and run the graph. An edited picture comes back, much like the before and after pair the card shows.

For newcomers and experts: ComfyUI is a tool where you connect blocks to run an image model on your own computer. Fibo 2.0 is an image-editing model from BRIA AI, and these files are its form for ComfyUI. A first small step is to download the files, place them and open one of the edit workflows.

The bf16 files combine the transformer shards from diffusers into a single file and rename the keys to match ComfyUI. Query, key and value weights are joined into one qkv tensor, and the cross-attention and Perceiver key and value weights into kv. Values are unchanged. The int8 files quantize the same weights with a Hadamard rotation and per-channel scales. The adaLN modulation layers stay in bf16, and outputs differ slightly from the original.

Download the files from the repository
Place them in the ComfyUI model folders
Open a Dev or Turbo edit workflow
Run it and get the edited picture

Good to know:

• The weights belong to BRIA AI. The license allows non-commercial use only, and commercial use needs a separate agreement with BRIA AI.
• The weights come as is, without any warranty.
• The int8 versions give slightly different outputs from the original model.
• Another change fuses two feed-forward weights into one and pads the hidden width to 9728.
• License: cc-by-nc-4.0.
• Cost to run: none named, you use your own hardware.
• Data it reaches: files downloaded from Hugging Face, plus the pictures you edit on your own machine.
• Not for: commercial use without a separate agreement with BRIA AI.

On Hugging Face, 13 November 2026

• Task: image to image
• Library: diffusion-single-file
• Downloads in the last 30 days: 407,328
• Likes: 232
• License: cc-by-nc-4.0
• Based on: briaai/Fibo-2.0-edit-turbo
• Quantized versions: 1
• Tags: diffusion-single-file, comfyui, image-to-image

Links

• Model: https://huggingface.co/Comfy-Org/Fibo-2.0
• Files: https://huggingface.co/Comfy-Org/Fibo-2.0/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download Comfy-Org/Fibo-2.0

Then follow the usage in its model card. Nothing in this kit runs by itself.