Fix README: correct repo id, ONNX vs PT split, GitHub link

README.md

@@ -1,26 +1,51 @@
 ---
 license: mit
+language:
+  - multilingual
+tags:
+  - text-to-speech
+  - speech-synthesis
+  - hebrew
+pipeline_tag: text-to-speech
 ---
 
-# Blue Model Checkpoints
+# Blue — PyTorch weights (training, finetuning & voice export)
 
-This repository contains the training checkpoints and stats files for the [Blue](https://github.com/maxmelichov/BlueTTS) speech synthesis system.
+This repository contains **Safetensors / PyTorch checkpoints** and **multilingual latent statistics** for **[BlueTTS](https://github.com/maxmelichov/BlueTTS)** — Hebrew-first multilingual text-to-speech with optional English, Spanish, Italian, German, and mixed-language synthesis in the reference code.
 
-## Checkpoints Directory
+**Project home (install, ONNX inference, examples):** [https://github.com/maxmelichov/BlueTTS](https://github.com/maxmelichov/BlueTTS)
 
-If you are following the finetuning or training instructions, download these weights:
+> **End-user synthesis:** Use the ONNX model bundle **[`notmax123/blue-onnx`](https://huggingface.co/notmax123/blue-onnx)** with the BlueTTS README. This **`notmax123/blue`** repo supplies **training / finetuning weights** and files needed to **export new voice style JSON** for ONNX; it is not the ONNX runtime bundle.
 
-- `blue_codec.safetensors`: The standalone trained Audio Codec. Used for translating to/from discrete/continuous latents.
-- `stats_multilingual.pt`: The statistical data containing the latent means and standard deviations computed from the corpus.
-- `vf_estimator.safetensors`: The combined Text-to-Latent acoustic checkpoints (includes text encoder, reference encoder, and the Flow Matching model).
-- `duration_predictor.safetensors`: The Duration Predictor checkpoint.
+## Files
 
-## Setup Instructions
+| File | Role |
+|------|------|
+| `blue_codec.safetensors` | Audio codec: mel ↔ latent, discrete/continuous conversion. |
+| `stats_multilingual.pt` | Latent mean/std for normalization (same statistics as training). |
+| `vf_estimator.safetensors` | Text-to-latent acoustic model (text encoder, reference encoder, flow-matching core). |
+| `duration_predictor.safetensors` | Duration predictor checkpoint. |
 
-To train or finetune, download this repository to your codebase:
+## Download
+
+Repo id is **case-sensitive** — use `notmax123/blue` (not `Blue`).
+
+```bash
+hf download notmax123/blue --repo-type model --local-dir ./pt_weights
+```
+
+Equivalent with the classic CLI:
 
 ```bash
-huggingface-cli download notmax123/Blue --local-dir pt_weights
+huggingface-cli download notmax123/blue --repo-type model --local-dir ./pt_weights
 ```
 
-For more specifics, see the [Training Guide](https://github.com/maxmelichov/BlueTTS/tree/main/training).
+## How to use
+
+1. **Training or finetuning:** Follow the [training](https://github.com/maxmelichov/BlueTTS/tree/main/training) directory in the BlueTTS GitHub repository.
+
+2. **New voices for ONNX inference:** Clone [BlueTTS](https://github.com/maxmelichov/BlueTTS), install with the `export` extra, download these weights locally, and run `scripts/export_new_voice.py` (see script docstring and project README).
+
+## License
+
+MIT — see the [BlueTTS repository](https://github.com/maxmelichov/BlueTTS) for the full license text.