A collection of INT4 ConvRot-quantized diffusion, video, and upscaling models for ComfyUI, built to run comfortably on 8GB-class GPUs (developed and tested on an RTX 3070 Ti) without gutting output quality.

What's in this repo

Model	Type	Notes	Quant
Flux 2 Dev	Image diffusion	Base Flux 2 Dev checkpoint	INT4 convrot
Gemma 3 12B IT	LLM / Text	Instruction-tuned language model	INT4 convrot
Ideogram 4 Instant Comfy	Image diffusion	Comfy-optimized Ideogram 4 variant	INT8 convrot
Krea 2 Raw	Image diffusion	Base Krea 2 checkpoint, unmodified pipeline	INT4 convrot
Krea 2 Turbo	Image diffusion	Distilled/turbo variant, fewer steps	INT4 convrot
LTX-2.3 22B Distilled 1.1	Video diffusion	Transformer-only distilled LTX-2.3 build (22B)	INT4 convrot
Mistral 3 Small Flux 2	LLM / Multimodal	Mistral 3 Small integrated with Flux 2	INT4 convrot
SAM 3.1 Multiplex	Vision / Segmentation	Segment Anything Model 3.1 multiplex variant	INT8 convrot
SeedVR2 (7B)	Image and Video Upscaler	Full 7B variant	INT4 convrot
Sulphur Distilled	Video diffusion	Distilled checkpoint built off of LTX-2.3	INT4 convrot

All models are quantized to INT4 or INT8 using the Starnodes Model Converter (https://github.com/Starnodes2024/comfyui-starnodes-modelconverter).

Why ConvRot INT4

Standard INT8/INT4 row-wise quantization throws away a lot of precision on the weight matrices that matter most for visual fidelity. ConvRot groups weights along their largest power-of-4-compatible dimension before quantizing, which keeps much more of the original model's detail and reduces the artifacting you'd normally see from a naive INT4 cast. The trade-off is VRAM and disk savings big enough to run models like SeedVR2 7B and full video diffusion checkpoints on 8GB cards.

Requirements

ComfyUI (nighlty build)
If you're getting chronic errors update your Conda environment (I'm running Pytorch 2.12, cu132, Python 3.12, Flashattention/Sageattention and Triton 3.8)

Installation

Install ComfyUI-INT4-Fast into ComfyUI/custom_nodes/
Download the model(s) you want from this repo into the matching ComfyUI/models/diffusion_models/ (or appropriate folder for video/upscale models)
Load with the INT4 loader node from ComfyUI-INT4-Fast — do not use the standard checkpoint/UNETLoader nodes, they will not decode these correctly
See the Samples and Workflow folder in this repo for ready-to-use ComfyUI workflow JSONs and sample outputs

Quantization pipeline

Built with Starnodes power:

Grab the Starnodes model converter and do it yourself if you wish. It supports INT8 and INT4 convrot: Starnodes

License

Inherits the license terms of each respective base model (Krea 2, LTX-2.3, Sulphur 2, SeedVR2). Check each upstream model's license before commercial use.

Part of the ⚡ Winnougan quantization series.

Ideogram Non-Commercial Model Agreement Summary

Last Updated: June 3, 2026
Model Covered: Ideogram 4

Disclaimer: This is a plain-English summary for informational purposes only and does not constitute legal advice. Always refer to the original legal text for binding obligations.

🎯 Core Purpose

This is a strictly non-commercial license. It allows you to use, modify, and share the Ideogram 4 model and its derivatives, but only for non-commercial, research, personal, or internal testing purposes.

✅ What You CAN Do

Use & Modify: Use, copy, modify, and create derivative works (e.g., fine-tuned versions) of the model.
Permitted Non-Commercial Uses:
- Personal research, hobbies, or private entertainment.
- Internal testing, evaluation, or R&D by for-profit companies (provided it is not in a live, customer-facing, or production environment).
- Use by charitable organizations for charitable purposes.
Ownership: You own the incremental rights to any Model Derivatives you create. Ideogram claims no ownership over the Outputs (e.g., images, text) you generate.

❌ What You CANNOT Do

No Commercial Use: You cannot use the model or its outputs to directly or indirectly generate revenue, or include them in revenue-generating products/services.
No Competing Models: You cannot use the model’s outputs to train, fine-tune, or distill a competing AI model or product.
Prohibited High-Risk/Illegal Uses:
- Military or surveillance purposes (including related R&D).
- Biometric processing.
- Generating unlawful content (e.g., CSAM, non-consensual intimate images).
- Making automated decisions in high-risk domains (finance, legal, employment, healthcare, housing, insurance, social welfare).
- Violating privacy/data protection laws or third-party rights (including rights of publicity).
No Circumvention: You cannot remove, bypass, or disable Ideogram’s built-in safety measures, content filters, or watermarking.

📦 Rules for Redistribution (Sharing the Model)

If you share or distribute the model or your derivatives, you must:

Apply the same non-commercial restrictions to anyone you share it with.
Provide all third-party recipients with a copy of this Agreement.
Include the following exact attribution notice in a "Notice" text file accompanying the copy:

"Ideogram 4 is provided under and subject to the Ideogram Non-Commercial Model Agreement available at https://github.com/ideogram-oss/ideogram-4/model_licenses/LICENSE-IDEOGRAM-4-NON-COMMERCIAL. All rights reserved. Copyright © Ideogram, Inc."
Clearly mark any files you have modified with prominent notices.
Not misrepresent your derivative as an official, endorsed, or validated Ideogram product.

⚖️ Liability, Warranties, and Legal Terms

"As Is" Basis: The model is provided without any warranties (express or implied). You assume all risks associated with using, modifying, or distributing it.
Limitation of Liability: Ideogram is not liable for any damages (direct, indirect, special, consequential, or punitive) arising from your use of the model.
Indemnification: You agree to cover Ideogram’s legal costs, losses, and damages if a third party sues them because of how you used, distributed, or breached the agreement regarding the model or its outputs.
Termination: Ideogram can terminate this agreement at any time, or immediately if you breach the terms (or if you sue them over IP rights). Upon termination, you must delete the model and cease all use/distribution.
Governing Law: The agreement is governed by the laws of the State of New York, without regard to conflicts of law provisions.

💡 Key Takeaway

You are free to experiment, research, and build non-commercial projects with Ideogram 4, provided you give proper attribution, keep it strictly non-commercial, do not use it to train competitors, and avoid harmful or high-risk applications. For commercial use, a separate agreement with Ideogram is required.

Winnougan/INT4-Convrot-Comfy-Models

INT4 ConvRot Comfy Models — Winnougan