Sign In

Fibo 2.0 Edit Turbo 12GB — int8 Image Editing in 4 steps + Qwen3-VL Prompt Enhancer

Download

1 variant available

Config Other

The_frizzy1_fibo2-edit-turbo-12gb_v1.0.0.json

53.84 KB

Verified:

Type
Workflows
Stats

6

Reviews

No reviews yet

Published

Nov 20, 2026

Base Model

Other

Hash
AutoV24A3A5CBC1D
01_fibo_edit_pe_20261028.png

☕ One coffee = one happy creator: https://buymeacoffee.com/the_frizzy1


No video on this workflow yet — the one above is how to use my workflows, if you are new to them.

Fibo 2.0 Edit Turbo on a 12 GB card: say what the picture should look like, the prompt enhancer writes Fibo's structured JSON caption, 4 steps later the edit is done. One workflow file, core ComfyUI nodes only. Write what the result looks like, not what stays the same.

What you need

✔ Custom nodes — None. Core ComfyUI only.
✔ ComfyUI version — 0.43 or newer. 0.43.1 tested, with sage attention.
✔ VRAM — 12 GB — tested, not estimated. Peak 11.6 GB, whole card as nvidia-smi reports it. No headroom.
✔ System RAM — 3.4 GB peak for ComfyUI, in a 32 GB machine
✔ Tested on — RTX 3060 12 GB · 32 GB RAM · ComfyUI 0.43.1 · sage attention
✔ Licence — Workflow MIT · Fibo 2.0: CC BY-NC 4.0, non-commercial only · Qwen3-VL: Apache-2.0

Models to download

Click a file name to download it. Put it in the folder shown, inside ComfyUI/models/.

Required — 4 files, 23.0 GB. If you already run FLUX 3 Klein, two of these are files you have.

diffusion_models/

File: fibo2_edit_turbo_int8_convrot.safetensors — 8.41 GB (edit)
Download:
https://huggingface.co/Comfy-Org/Fibo-2.0/resolve/main/diffusion_models/fibo2_edit_turbo_int8_convrot.safetensors

Put it here:
ComfyUI/models/diffusion_models/

text_encoders/

File: qwen3vl_4b_int8_convrot.safetensors — 4.86 GB
Download:
https://huggingface.co/Comfy-Org/Fibo-2.0/resolve/main/text_encoders/qwen3vl_4b_int8_convrot.safetensors

Put it here:
ComfyUI/models/text_encoders/

File: qwen3vl_8b_int8_convrot.safetensors — 9.35 GB (prompt enhancer)
Download:
https://huggingface.co/Comfy-Org/Qwen3-VL/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors

Put it here:
ComfyUI/models/text_encoders/

vae/

File: flux2-vae.safetensors — 0.34 GB
Download:
https://huggingface.co/Comfy-Org/Fibo-2.0/resolve/main/vae/flux2-vae.safetensors

Put it here:
ComfyUI/models/vae/

Lighter than the official template: Qwen3-VL-4B int8 as the text encoder instead of bf16.

Speed on my RTX 3060

Wall is queue to finished. VRAM is the whole card as nvidia-smi reports it, about 0.7 GB of that is Windows.

  • Edit, first run after starting ComfyUI (loads Fibo + enhancer) — about 1 MP, 1 min 9 s wall, 11.3 GB peak vram, 3.2 GB peak ram (comfyui)

  • Edit, new prompt, enhancer on — about 1 MP, 1 min 6 s wall, 11.4 GB peak vram, 3.4 GB peak ram (comfyui)

  • Edit, enhancer off (the wrong result, see below) — about 1 MP, 15 s wall, 11.6 GB peak vram, 3.2 GB peak ram (comfyui)

Most of the time is the enhancer. The honest number is about one minute per edit — the 15 s run is the broken one.

What did not work

  • The enhancer is the slow part and you still want it on. Fibo reads a structured JSON caption, not a sentence. I switched it off once and gave Fibo the plain sentence "The net is bright blue.": the fisherman got a white blazer, red curly hair and sunglasses, and the net stayed red. Off only makes sense if you write the JSON yourself.

  • Peak sits at 11.6 GB on a 12 GB card. There is no headroom.

  • Only three runs were measured, all at about 1 MP. Nothing bigger was tried, so I cannot tell you what a 2K edit costs.

  • The licence is the other catch: Fibo 2.0 is CC BY-NC 4.0, non-commercial only, so nothing you make with it goes in a paid job.

Get the workflow

More videos on this stuff: https://www.youtube.com/playlist?list=UUGKDBMhTGqvlBZE9cF5d50Q