README.md

---
title: Anima · Telescopa
emoji: 🎨
colorFrom: indigo
colorTo: pink
sdk: gradio
sdk_version: 6.14.0
app_file: app.py
pinned: false
models:
  - RicemanT/Anima-Telescopa
  - circlestone-labs/Anima
---

# Anima · Telescopa

A Gradio demo for [`RicemanT/Anima-Telescopa`](https://huggingface.co/RicemanT/Anima-Telescopa), a full-matrix **LyCORIS LoKr** fine-tune of [`circlestone-labs/Anima`](https://huggingface.co/circlestone-labs/Anima).

## What this Space does

- Downloads Anima's split diffusion model, Qwen text encoder, VAE, and tokenizers from the Hub.
- Applies one of the published Telescopa LoKr variants at runtime.
- Generates anime-focused text-to-image illustrations with prompt, negative prompt, resolution, seed, LoKr strength, CFG, steps, and AuraFlow shift controls.
- Keeps the base model and adapter in the Hugging Face cache so later generations do not repeat downloads.

The first generation is intentionally slower because the base model and selected adapter are large. A GPU-enabled Space hardware tier is required for practical inference.

## Run locally

```bash
cd anima-telescopa-space
python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt
python app.py
```

A local CUDA GPU with enough VRAM is strongly recommended. The application starts without a GPU so the interface can be inspected, but generation reports a clear GPU requirement instead of attempting an impractical CPU run.

## Deploy to Hugging Face Spaces

Create a new **Gradio** Space and copy the contents of this directory into it, or push this directory to a Space repository after authenticating with the Hugging Face CLI. The README frontmatter already sets the app entry point and model references.

## License

`RicemanT/Anima-Telescopa` is published under the **CircleStone Labs Non-Commercial License v1.1**. Review the [upstream model card and license](https://huggingface.co/RicemanT/Anima-Telescopa) before deploying the demo or using its weights. The base Anima model has its own license terms as well.