Uploaded October 2026 | Updated November 2026, 2 days ago
I tested Bria FIBO 2.0 in ComfyUI to see how Dev and Edit Turbo handle image generation, editing, and VGL JSON prompts. You'll see the settings I tried, the results I got, and the problems I found.
I compared Edit Turbo at 4 and 8 steps with CFG 1, and Dev at 30 and 50 steps with CFG 5. I also tried generating images with Edit Turbo using only a text prompt, even though it is mainly designed for editing.
One of the most interesting parts is VGL (Visual GenAI Language). I tested normal prompts against structured JSON prompts to see how well FIBO follows object counts, camera angles, lighting, and other details.
The JSON did not always work on the first try. In my fruit test, I had to correct the JSON to get five apples and one pear. I also changed individual fields to see how the model handled lighting, plate colour, and camera angle.
For image editing, I tested changing a jacket, relighting a photo, and taking clothing from a second reference image. I show why clear image roles matter when you want to keep the same face and pose.
Then I used four images to build one scene: a street, a tank, a checkpoint, and a fire. You can see how the result changes as I add each reference image.
I also tested text generation, including a shop sign, a menu with prices, and Hindi text. The English text worked in my examples, but Hindi failed.
**FIBO 2.0 model files and resources**
Workflow + VGL Template: patreon.com/ComfyuiWorkflowBlog/posts/fibo-2-0-comfyui-173198107
ComfyUI-ready model files:
huggingface.co/Comfy-Org/Fibo-2.0
FIBO 2.0 Dev:
huggingface.co/briaai/Fibo-2.0-dev
FIBO 2.0 Edit Turbo:
huggingface.co/briaai/Fibo-2.0-edit-turbo
**License:** The linked model weights are for non-commercial use. Check Bria's licensing terms before using them for client or commercial work:
platform.bria.ai/licensing/pricing?model=fibo-2
If you try FIBO 2.0 with Hindi or another language, tell me in the comments what worked and what failed. I would like to test more languages.
Subscribe for more ComfyUI tutorials, AI model tests, and image editing workflows.
#FIBO2 #ComfyUI #AIImageEditing
I tested Bria FIBO 2.0 in ComfyUI to see how Dev and Edit Turbo handle image generation, editing, and VGL JSON prompts. You'll see the settings I tried, the results I got, and the problems I found.
I compared Edit Turbo at 4 and 8 steps with CFG 1, and Dev at 30 and 50 steps with CFG 5. I also tried generating images with Edit Turbo using only a text prompt, even though it is mainly designed for editing.
One of the most interesting parts is VGL (Visual GenAI Language). I tested normal prompts against structured JSON prompts to see how well FIBO follows object counts, camera angles, lighting, and other details.
The JSON did not always work on the first try. In my fruit test, I had to correct the JSON to get five apples and one pear. I also changed individual fields to see how the model handled lighting, plate colour, and camera angle.
For image editing, I tested changing a jacket, relighting a photo, and taking clothing from a second reference image. I show why clear image roles matter when you want to keep the same face and pose.
Then I used four images to build one scene: a street, a tank, a checkpoint, and a fire. You can see how the result changes as I add each reference image.
I also tested text generation, including a shop sign, a menu with prices, and Hindi text. The English text worked in my examples, but Hindi failed.
**FIBO 2.0 model files and resources**
Workflow + VGL Template: patreon.com/ComfyuiWorkflowBlog/posts/fibo-2-0-comfyui-173198107
ComfyUI-ready model files:
huggingface.co/Comfy-Org/Fibo-2.0
FIBO 2.0 Dev:
huggingface.co/briaai/Fibo-2.0-dev
FIBO 2.0 Edit Turbo:
huggingface.co/briaai/Fibo-2.0-edit-turbo
**License:** The linked model weights are for non-commercial use. Check Bria's licensing terms before using them for client or commercial work:
platform.bria.ai/licensing/pricing?model=fibo-2
If you try FIBO 2.0 with Hindi or another language, tell me in the comments what worked and what failed. I would like to test more languages.
Subscribe for more ComfyUI tutorials, AI model tests, and image editing workflows.
#FIBO2 #ComfyUI #AIImageEditing






![FLUX.2 Klein ComfyUI Workflow: Image Edit New Trick & ControlNet
If your FLUX.2 Klein results look like plastic or your GPU is struggling, you have to watch this. I created a new Flux Klein ComfyUI Workflow that fixes the realism issues and runs on low VRAM. I also found a secret ControlNet Hack, image editing trick that nobody know
In this video, I break down exactly how to use the new NVFP4 model format. It cuts VRAM usage by 55 percent, making it perfect for 4090 and 5090 cards. If you have an 8GB card, I show you how to use the GGUF model to get the same high quality.
You will also learn how to unlock the hidden Beta 57 scheduler by installing the RES4LYF node, which is the key to fixing that plastic AI look. Finally, I show you my dedicated image editing workflow to change clothes and transfer poses perfectly.
DFans – AI Influencer Monetization Platform
https://dfans.co/?ref=LA63HR
invitation code:- LA63HR
Files And Workflows You Need To Download:- https://aistudynow.com/flux-2-klein-fix-crashes-run-9b-on-6gb-vram-workflow-download/
Monetize Your ComfyUI Content [Link to your monetization offer]
#fluxklein #comfyui #comfyuiworkflow FLUX.2 Klein ComfyUI Workflow: Image Edit New Trick & ControlNet](https://i.ytimg.com/vi/jV7SNCMEKMw/mqdefault.jpg)



