README.md

metadata

title: Anima
emoji: 🎨
colorFrom: purple
colorTo: pink
sdk: docker
pinned: false

Anima Preview3 — ComfyUI Space

Anime-focused text-to-image generation powered by Anima Preview3 (2B params) running on CPU via ComfyUI.

Usage

The ComfyUI interface is available directly in this Space. Drag in the workflow from the Anima model card to get started.

API

ComfyUI's REST API is exposed on the same port as the UI:

Endpoint	Method	Description
/prompt	POST	Submit a generation job (send workflow JSON)
/queue	GET	View current queue
/history/{prompt_id}	GET	Poll job status + get output filename
/view?filename=...	GET	Fetch the output image

Recommended prompting

masterpiece, best quality, score_7, safe, [your tags here]

Negative: worst quality, low quality, score_1, score_2, score_3

Use the Turbo LoRA (already loaded) with 14 steps for much faster generation on CPU.

Notes

• Runs fully on CPU — generation takes a few minutes per image
• Optimized with BF16, jemalloc, and OpenMP thread tuning
• Based on rikunarita/Anima-Preview-2 with Preview3 weights