SeeSee21's Z-Anime Brings Alibaba's 6B DiT to Anime Generation
SeeSee21 released Z-Anime, a full fine-tune of Alibaba's Z-Image on anime data, running on 8GB VRAM with 4-step distilled variants
PRO- Z-Anime is a full fine-tune (not a LoRA merge) of Alibaba's Z-Image Base for anime generation.
- Built on S3-DiT, a 6B-parameter Single-Stream Diffusion Transformer with Qwen3-4B as the text encoder.
- Ships as Base, Distill-8-Step, and Distill-4-Step variants in BF16, FP8, GGUF, and AIO formats.
- Runs on 8GB VRAM; GGUF Q4_K_S weighs ~4.2GB for CPU or low-VRAM setups.
- Uses natural language prompts, not Danbooru-style tags; euler_ancestral + beta scheduler recommended.
- Apache 2.0 licensed, partially NSFW capable, and available at SeeSee21/Z-Anime.
A new anime-focused checkpoint just landed on Hugging Face, and it takes a slightly different route than the usual Illustrious or Pony spinoffs. Z-Anime from SeeSee21 is a full fine-tune of Alibaba's Z-Image Base rather than a LoRA merge, and it ships as a family of variants tuned for quality, speed, or low-VRAM inference.
Z-Image itself is Alibaba Tongyi Lab's open text-to-image model built on the S3-DiT (Single-Stream Diffusion Transformer) architecture with 6B parameters. It covers a broad aesthetic spectrum from hyper-realistic photography to anime and stylized illustration, with high variability in composition, facial identity, and lighting across seeds. SeeSee21 has taken that base and specialized it entirely for anime generation.
Fine-tune, not a LoRA merge
The distinction matters. Z-Anime is trained from the ground up on the S3-DiT backbone, inheriting the parent model's diversity, controllability, full negative prompt support, and headroom for further fine-tuning. Anime aesthetics are baked into the weights themselves rather than layered on top through adapter deltas, which usually translates into cleaner style consistency and a better base for stacking your own LoRAs.
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
