Update README.md

README.md

@@ -1,119 +1,10 @@
-# LEMAS-TTS Gradio Demo (Hugging Face Space)
-
-This folder is a **clean, inference-only** version of LEMAS-TTS, organized for easy deployment on **Hugging Face Spaces**.
-
-It keeps only:
-- the inference models & configs (`lemas_tts`)
-- pretrained checkpoints and vocab (`pretrained_models`)
-- the bundled UVR5 denoiser (`uvr5`)
-- a Gradio web UI (`inference_gradio.py`, `app.py`)
-
 ---
-
-## 1. Features
-
-- Zero-shot TTS: clone voice from a reference audio + reference text
-- Multilingual text input (Chinese / English / ES / IT / PT / DE, etc.)
-- Optional UVR5-based reference denoising
-- Two custom LEMAS checkpoints:
-  - `multilingual_prosody_custom`
-  - `multilingual_acc_grl_custom`
-
+title: LEMAS-Edit
+emoji: ✨
+colorFrom: indigo
+colorTo: green
+sdk: gradio
+sdk_version: "5.10.0"
+app_file: app.py
+pinned: false
 ---
-
-## 2. Project Structure
-
-```text
-LEMAS-TTS_gradio/
-  app.py                     # HF Space entrypoint (Gradio Blocks)
-  inference_gradio.py        # Full Gradio UI & logic
-  requirements.txt           # Minimal runtime dependencies
-
-  lemas_tts/                 # Core LEMAS-TTS package (inference only)
-    api.py                   # F5TTS API (used by the UI)
-    configs/                 # Model configs (F5TTS / E2TTS)
-    infer/                   # Inference utilities & text frontend
-    model/                   # DiT backbone, utils, etc.
-
-  pretrained_models/         # All local assets needed for inference
-    ckpts/
-      F5TTS_v1_Base_vocos_custom_multilingual_prosody/model_2698000.pt
-      F5TTS_v1_Base_vocos_custom_multilingual_acc_grl/model_2680000.pt
-      prosody_encoder/...
-      vocos-mel-24khz/...
-    data/
-      multilingual_prosody_custom/vocab.txt
-      multilingual_acc_grl_custom/vocab.txt
-      test_examples/*.wav    # Demo audios used in the UI
-    uvr5/
-      models/MDX_Net_Models/model_data/*.onnx, *.json
-
-  uvr5/                      # Bundled UVR5 implementation for denoising
-```
-
-`lemas_tts.api.F5TTS` automatically resolves `pretrained_models/` based on the repo layout, so no extra path configuration is required.
-
----
-
-## 3. How to Run Locally
-
-```bash
-cd LEMAS-TTS_gradio
-pip install -r requirements.txt
-python app.py
-```
-
-Then open the printed URL (default `http://127.0.0.1:7860`) in your browser.
-
----
-
-## 4. Hugging Face Space Setup
-
-1. Create a new Space (type: **Gradio**).
-2. Upload the contents of `LEMAS-TTS_gradio/` to the Space repo:
-   - `app.py`
-   - `inference_gradio.py`
-   - `requirements.txt`
-   - `lemas_tts/`
-   - `pretrained_models/`
-   - `uvr5/`
-3. In the Space settings, choose a GPU hardware profile (the model is heavy).
-4. The Space will automatically run `app.py` and launch the Gradio Blocks named `app`.
-
-No extra arguments are needed; all paths are relative inside the repo.
-
----
-
-## 5. Usage Tips
-
-- **Reference Text** should match the reference audio roughly in content and language for best voice cloning.
-- **Denoise**:
-  - Turn on if your reference audio is noisy; it runs UVR5 on CPU.
-  - Turn off if the reference is already clean (saves time).
-- **Seed**:
-  - `-1` → random seed
-  - Any other integer → reproducible output
-
----
-
-## 6. 中文说明（简要）
-
-这个目录是专门为 **Hugging Face Space** 打包的 **推理版 LEMAS-TTS**：
-
-- 只保留推理相关代码（`lemas_tts`）、预训练模型（`pretrained_models`）和 UVR5 去噪模块（`uvr5`）
-- Gradio 入口为 `app.py`，内部调用 `inference_gradio.py` 里的 `app`（一个 `gr.Blocks` 界面）
-- `pretrained_models/` 下已经包含：
-  - 自定义多语种 prosody / accent GRL 的 finetune 权重
-  - vocoder（`vocos-mel-24khz`）
-  - prosody encoder
-  - 以及示例语音 `test_examples/*.wav`
-
-在本地或 Space 中运行步骤：
-
-```bash
-pip install -r requirements.txt
-python app.py
-```
-
-然后在浏览器中打开提示的链接即可使用零样本 TTS Demo。
-