Free ComfyUI Colab Pack

Free Google Colab notebooks for popular ComfyUI workflows on low VRAM GPUs (focused on Colab Free T4).

Why this project

I put a lot of time and effort into these notebooks. They are tuned, tested, and maintained so people can run popular models for free in Colab.

The goal is simple: fast, practical, and stable model access in Colab without heavy local setup.

Killer Features

• Auto quant selection by VRAM budget (rare in public Colab packs).
• GGUF-first setup for low VRAM use cases.
• Match-only Lenovo UltraReal LoRA download by base-model tag.
• Built-in Hugging Face + Civitai token prompts.
• Stable Cloudflare tunnel launch logic with retries and health checks.
• Low-VRAM defaults for Colab T4 (memory-aware settings).
• Ready-to-use per-model workflow JSON files in this repo.

Notebook Catalog

Flux SRPO

• What: FLUX-based SRPO GGUF text-to-image notebook tuned for quick T4 runs.
• Model creators/sources: FLUX.1 family by Black Forest Labs, SRPO model by Tencent Hunyuan (tencent/SRPO), GGUF conversion pack by befox.
• Workflow: workflows/flux_srpo/flux_schnell_full_text_to_image.json
• Preview image: coming soon.

Flux2 Klein 9B GGUF

• What: Flux.2 Klein 9B base/distilled GGUF notebook for T2I and edit flows.
• Model creators/sources: FLUX.2 family by Black Forest Labs, GGUF releases by unsloth, VAE package by Comfy-Org.
• Workflows: workflows/flux2_klein9b_gguf/
• Preview image: coming soon.

Z-Image Base

• What: Z-Image base GGUF setup for quality-oriented generation.
• Model creators/sources: original Z-Image by Tongyi-MAI (Tongyi-MAI/Z-Image), GGUF ports by unsloth, ComfyUI split assets used from Comfy-Org/z_image.
• Workflow: workflows/zimage_base/zimage_base_test_workflow.json
• Preview image: coming soon.

Z-Image Turbo

• What: fast Z-Image Turbo GGUF notebook for speed-first generation.
• Model creators/sources: original Z-Image Turbo by Tongyi-MAI (Tongyi-MAI/Z-Image-Turbo), GGUF ports by unsloth, ComfyUI split assets used from Comfy-Org/z_image.
• Workflow: workflows/zimage_turbo/zimage_turbo_test_workflow.json
• Preview image: coming soon.

Z-Image Turbo + Base

• What: combo notebook with Turbo + Base variants in one setup.
• Model creators/sources: original Z-Image models by Tongyi-MAI (Tongyi-MAI/Z-Image and Tongyi-MAI/Z-Image-Turbo), GGUF variants by unsloth, ComfyUI split assets used from Comfy-Org/z_image.
• Workflows: workflows/zimage_turbo_base/
• Preview image: coming soon.

Z-Image Turbo + SeedVR2 Upscaler

• What: two-stage pipeline (Z-Image generation + SeedVR2 upscaling).
• Model creators/sources: original Z-Image Turbo by Tongyi-MAI (Tongyi-MAI/Z-Image-Turbo) with GGUF ports by unsloth, SeedVR2 node/files by numz and GGUF pack by cmeka.
• Workflows: workflows/zimage_seedvr2/
• Preview image: coming soon.

Qwen Image 2512

• What: Qwen Image 2512 GGUF generation notebook with optional Lightning LoRA.
• Model creators/sources: Qwen family by Alibaba/Qwen team, GGUF packs by unsloth and ggml-org, Lightning LoRA by lightx2v.
• Workflow: workflows/qwen_image_2512/qwen_image_test_workflow.json
• Preview image: coming soon.

Qwen Image Edit 2511

• What: Qwen Image Edit 2511 notebook for image editing use cases.
• Model creators/sources: Qwen family by Alibaba/Qwen team, GGUF packs by unsloth and ggml-org, Lightning LoRA by lightx2v.
• Workflow: workflows/qwen_image_edit_2511/qwen_image_edit_test_workflow_v2.json
• Preview image: coming soon.

Chroma1 HD GGUF

• What: Chroma1-HD text-to-image GGUF notebook.
• Model creators/sources: Chroma1-HD by lodestones, GGUF package by silveroxides.
• Workflow: workflows/chroma1_hd_gguf/ComfyUI_Chroma1-HD_T2I-workflow.json
• Preview image: coming soon.

Paused (Not In Active Testing)

• notebooks/_paused/ltx2_gguf/
• notebooks/_paused/wan22_14b_combo/

These are kept in repo and can be returned to active catalog after validation.

Repository Layout

free-comfyui-colab-pack/
  notebooks/<model>/comfy_<model>.ipynb
  notebooks/_paused/<model>/comfy_<model>.ipynb
  workflows/<model>/*.json
  workflows/_paused/<model>/*.json
  docs/

Quick Start

1. Open a notebook from notebooks/<model>/ in Colab.
2. Run cells top-to-bottom.
3. Enter Hugging Face token when asked.
4. Enter Civitai token when asked (recommended for LoRA downloads).
5. Open generated Cloudflare link and load workflow from workflows/<model>/.

Support

If these notebooks save you time, please support development:

I spend a lot of time on open-source projects. Even a $1 donation helps and goes directly to GPU servers and food.

Crypto Donations (Telegram Wallet)

Wallet addresses:

• TON: UQAMPvqduXVWyax325-zqk81rTwNG1bRhCvXPyIs7eeIxEVp
• USDT (TON): UQAMPvqduXVWyax325-zqk81rTwNG1bRhCvXPyIs7eeIxEVp
• Memo/Tag: check Wallet receive screen before sending.

See full support info in docs/SUPPORT.md.

Hire Me

• Email: ekkonwork@gmail.com
• Telegram: @Mikhail_ML_ComfyUI

Notes

• Respect model licenses and terms.
• Cloudflare tunnels can still be unstable in Colab due to external network conditions.