NitroFusion2 makes sharp text-to-image results in one step on top of Z-Image-Turbo

NitroFusion2Free open weights on Hugging Face (Apache 2.0); ComfyUI workflows on GitHub; Hugging Face demo.2026-11-03

NitroFusion2 is an open 1 to 2 step text-to-image model built on Z-Image-Turbo that keeps lettering and fine detail sharp even at one step. It ships with ComfyUI workflows and GGUF files from 2.0 GB to 7.2 GB.

What it does

A distilled text-to-image model that trains its adversarial loss on the text stream of the diffusion transformer, which the authors say helps typography and fine detail at one step. There are 1-step and 2-step ComfyUI workflows, GGUF versions of both, a diffusers script for 1 to 4 steps and an online demo.

Worth knowing

For 2 steps the authors recommend their de-distillation LoRA at 0.5 on the first step to remove some artifacts; the 2-step workflow already does this. Weights are Apache 2.0 and need the Z-Image-Turbo base pipeline. Paper on arXiv (2611.05648).

Try it yourself

Latina female with thick wavy hair, harbor boats and pastel houses behind. Breezy seaside light, warm tones, cinematic close-up.

Image Local

Read the primary source >

More coverage

Related entries

Open-H3-Regenerate-2K-IC regenerates a MiniMax H3 draft at 1080p, 2K or 4K while keeping its faces, lip sync and lettering
Open-H3-Regenerate-2K-IC, released on 9 November 2026 by the NVlabs SoL-Engine team, is an in-context LoRA for MiniMax H3 that regenerates a low-resolution draft at 1080p, 2K or 4K in 3, 4 or 8 steps. It is an open alternative to MiniMax's H3-Regenerate-2K, which is only a hosted API, and it keeps the draft in context so faces, lip sync, text and layout stay as they were.
Open-H3-Regenerate-2K2026-11-095 sources · huggingface.co
WeVisOne-VL now runs in ComfyUI core: one node turns a text prompt into boxes, masks, alpha mattes, points, counts or OCR
Tencent WeChat Vision's WeVisOne-VL, an Apache-2.0 vision model in 2B and 4B sizes, runs natively in ComfyUI since 9 November 2026, when kijai's pull request was merged. A single Detect node takes an image and a phrase and returns boxes, per-instance alpha mattes, masks, points, counts or OCR boxes, without SAM or custom nodes.
WeVisOne-VL2026-11-094 sources · github.com
ZONOS3-Mini is an open Apache-2.0 voice cloning text-to-speech model that runs in real time in a browser, on a laptop or on a phone
ZONOS3-Mini, released by Zyphra on 9 November 2026, is a compact open text-to-speech model with voice cloning and voice design, under the Apache 2.0 licence. It has 800 million parameters with about 100 million active per token, and runs in real time in a browser through WebGPU, on laptops and on phones.
ZONOS3-Mini2026-11-094 sources · huggingface.co
Fizgig 9.2 trains a MOSS-Image LoRA in about 20 minutes instead of 49 on an RTX 5090, with far less memory
Fizgig is a free, Apache-2.0 LoRA and fine-tune studio for image, video and music models such as Krea 2, MOSS-Image, Qwen Image 2.1, FLUX Klein, MiniMax H3 and ACE-Step. Version 9.2, released on 8 November 2026, moves to PyTorch 2.15 and a faster Muon optimizer step, so a MOSS-Image LoRA on an RTX 5090 trains in about 20 minutes instead of 49 with likeness unchanged.
Fizgig2026-11-083 sources · github.com
SnapRef v2a gives the Anima image model a character from one or more reference photos, with a weight for each photo, in ComfyUI
SnapRef is a reference adapter for the Anima anime image model: you give it one or more pictures of a character and the prompt only describes the scene, while hair, eyes and outfit come from the reference. The stable v2a release of 8 November 2026 adds a weight for each reference, two modes and six ComfyUI workflows.
SnapRef2026-11-083 sources · huggingface.co

< Back to all entries

NitroFusion2 makes sharp text-to-image results in one step on top of Z-Image-Turbo

What it does

Worth knowing

Try it yourself

More coverage

Related entries

Open-H3-Regenerate-2K-IC regenerates a MiniMax H3 draft at 1080p, 2K or 4K while keeping its faces, lip sync and lettering

WeVisOne-VL now runs in ComfyUI core: one node turns a text prompt into boxes, masks, alpha mattes, points, counts or OCR

ZONOS3-Mini is an open Apache-2.0 voice cloning text-to-speech model that runs in real time in a browser, on a laptop or on a phone

Fizgig 9.2 trains a MOSS-Image LoRA in about 20 minutes instead of 49 on an RTX 5090, with far less memory

SnapRef v2a gives the Anima image model a character from one or more reference photos, with a weight for each photo, in ComfyUI