One- and two-step image generation model distilled from Z-Image-Turbo using text-stream adversarial distillation.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
NitroFusion2 is a few-step text-to-image diffusion model from Dar-Yen Chen and Yi-Zhe Song, built on Z-Image-Turbo. It follows their earlier NitroFusion and applies adversarial diffusion distillation to a multimodal diffusion transformer (MMDiT), reducing generation from the usual 4–8 steps to 1–2 steps at comparable quality.
The method, Text-Stream Adversarial Distillation (TSAD), uses intermediate text-token states of a frozen MMDiT as the discriminator representation, since the text stream becomes image-dependent through joint attention. Aspect-Cued Discrimination (ACD) conditions the discriminator on aspects such as typography, objects or style to steer what the adversarial training emphasizes. The authors report sharp rendered text and fine detail at one step, and competitive results at two steps across evaluated benchmarks.
According to the project page and a YouTube creator summary, weights, ComfyUI workflows and a demo are available. The model is reported at about 12 GB in BF16, with compressed versions, and can be run in ComfyUI or Diffusers. The license was not stated in the provided material.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Image model based on Z-Image that generates in one or two diffusion steps. Model and code released, 12 GB BF16 plus compressed versions, runs in ComfyUI or Diffusers.
Closest neighbours by meaning, not just by category.
Research · ByteDance
ByteDance and Texas A&M research method distilling image and video diffusion models into few-step generators adversarially.
Model
Open-weights 6B text-to-image model family from Alibaba's Tongyi-MAI lab, with a fast 8-step Turbo variant.
Model · Princeton
Fully open 3B text-to-image diffusion model from Princeton with released weights, training code, and data pipeline.
Model
Pixel-space text-to-image diffusion model transferred from latent diffusion models, without a VAE, supporting 4K output.
Model
Text-to-image foundation model family (1B, 2B, 5B) using Semantic-First Diffusion with separate semantic and texture streams.
Model
Rank-asymmetric flow method for pixel-space image generation, incl. AsymFLUX.2 klein finetuned from FLUX.2 klein 9B.
