Update README.md

README.md

@@ -1,107 +1,111 @@
----
-title: Soprano 1.1 ONNX Web Demo
-emoji: 🎧
-colorFrom: blue
-colorTo: indigo
-sdk: static
-short_description: Real-time text-to-speech in the browser using ONNX
-app_file: index.html
-pinned: false
-models:
-- KevinAHM/soprano-1.1-onnx
-license: apache-2.0
----
-
-<!-- Version 0.0.3 -->
-<div align="center">
-
-# Soprano 1.1 ONNX Streaming — Instant Text‑to‑Speech in the Browser (WASM)
-
-[![Upstream](https://img.shields.io/badge/Upstream-ekwek1%2Fsoprano-black?logo=github)](https://github.com/ekwek1/soprano)
-[![Hugging Face Model](https://img.shields.io/badge/HuggingFace-Model-orange?logo=huggingface)](https://huggingface.co/KevinAHM/soprano-onnx)
-[![Hugging Face Demo for Soprano Web Onnx](https://img.shields.io/badge/HuggingFace-Demo-yellow?logo=huggingface)](https://huggingface.co/spaces/KevinAHM/soprano-web-onnx)
-
-A **static, client-side** browser demo that runs the Soprano TTS pipeline using **onnxruntime-web**. 
-
-Soprano 1.1 features significant performance optimizations, including moving all heavy inference to a **Web Worker** and utilizing an **int8 quantized decoder** for superior real-time speeds on consumer CPUs.
-
----
-
-## Requirements
-
-- A modern browser (Chrome, Edge, Firefox, Safari).
-- You must serve this folder over HTTP (opening `index.html` via `file://` usually breaks `fetch()` / module loading).
-- The demo loads `onnxruntime-web` and `@huggingface/transformers` from a CDN by default (network required unless you vendor them).
-- The model files are large; plan to use **Git LFS** or GitHub Releases if you publish them.
-
----
-
-## Folder layout
-
-Place model artifacts under `./models/`:
-
-```text
-.
-├─ index.html
-├─ onnx-streaming.js      (Main Thread Client)
-├─ inference-worker.js    (Heavy Inference Engine)
-├─ PCMPlayerWorklet.js    (Audio Playback Worklet)
-├─ style.css
-├─ onnx/
-│  ├─ soprano_backbone_kv_fp32.onnx
-│  └─ soprano_decoder_int8.onnx
-...
-```
-
-Notes:
-- ONNX models live in `onnx/` following HuggingFace convention.
-- The decoder uses external weights (`.onnx.data` file must be present alongside the `.onnx` file).
-- Tokenizer files are in the root directory.
-
----
-
-## Run locally
-
-Use any static file server from this directory, for example:
-
-```bash
-python -m http.server 8085
-```
-
-Then open `http://localhost:8085`.
-
----
-
-## Configuration
-
-Model paths are defined near the top of `onnx-streaming.js` in the `MODELS` object.
-
-Sampling defaults are set in `onnx-streaming.js` (constructor):
-- `temperature`
-- `topK`
-- `topP`
-- `repetitionPenalty`
-
----
-
-## Troubleshooting
-
-- **"Load failed" / model never becomes Ready**
-  - Verify the `onnx/` filenames match `MODELS` in `onnx-streaming.js`
-  - Check DevTools → Network for a missing `.onnx.data` file (404)
-  - Confirm `/` contains `tokenizer.json` (and related files)
-- **Performance notes**
-  - **Web Worker:** Keeps the UI responsive (no lag during generation).
-  - **int8 Decoder:** Optimized for high-throughput CPU inference.
-  - Achieves real-time streaming on modern hardware
-
----
-
-## License & attribution
-
-Soprano is released under **Apache-2.0** in the upstream repository:
-https://github.com/ekwek1/soprano
-
-
-
-
+---
+title: Soprano 1.1 ONNX Web Demo
+emoji: 🎧
+colorFrom: blue
+colorTo: indigo
+sdk: static
+short_description: Real-time text-to-speech in the browser using ONNX
+app_file: index.html
+pinned: false
+models:
+- KevinAHM/soprano-1.1-onnx
+license: apache-2.0
+custom_headers:
+  cross-origin-embedder-policy: require-corp
+  cross-origin-opener-policy: same-origin
+  cross-origin-resource-policy: cross-origin
+---
+
+<!-- Version 0.0.3 -->
+<div align="center">
+
+# Soprano 1.1 ONNX Streaming — Instant Text‑to‑Speech in the Browser (WASM)
+
+[![Upstream](https://img.shields.io/badge/Upstream-ekwek1%2Fsoprano-black?logo=github)](https://github.com/ekwek1/soprano)
+[![Hugging Face Model](https://img.shields.io/badge/HuggingFace-Model-orange?logo=huggingface)](https://huggingface.co/KevinAHM/soprano-onnx)
+[![Hugging Face Demo for Soprano Web Onnx](https://img.shields.io/badge/HuggingFace-Demo-yellow?logo=huggingface)](https://huggingface.co/spaces/KevinAHM/soprano-web-onnx)
+
+A **static, client-side** browser demo that runs the Soprano TTS pipeline using **onnxruntime-web**. 
+
+Soprano 1.1 features significant performance optimizations, including moving all heavy inference to a **Web Worker** and utilizing an **int8 quantized decoder** for superior real-time speeds on consumer CPUs.
+
+---
+
+## Requirements
+
+- A modern browser (Chrome, Edge, Firefox, Safari).
+- You must serve this folder over HTTP (opening `index.html` via `file://` usually breaks `fetch()` / module loading).
+- The demo loads `onnxruntime-web` and `@huggingface/transformers` from a CDN by default (network required unless you vendor them).
+- The model files are large; plan to use **Git LFS** or GitHub Releases if you publish them.
+
+---
+
+## Folder layout
+
+Place model artifacts under `./models/`:
+
+```text
+.
+├─ index.html
+├─ onnx-streaming.js      (Main Thread Client)
+├─ inference-worker.js    (Heavy Inference Engine)
+├─ PCMPlayerWorklet.js    (Audio Playback Worklet)
+├─ style.css
+├─ onnx/
+│  ├─ soprano_backbone_kv_fp32.onnx
+│  └─ soprano_decoder_int8.onnx
+...
+```
+
+Notes:
+- ONNX models live in `onnx/` following HuggingFace convention.
+- The decoder uses external weights (`.onnx.data` file must be present alongside the `.onnx` file).
+- Tokenizer files are in the root directory.
+
+---
+
+## Run locally
+
+Use any static file server from this directory, for example:
+
+```bash
+python -m http.server 8085
+```
+
+Then open `http://localhost:8085`.
+
+---
+
+## Configuration
+
+Model paths are defined near the top of `onnx-streaming.js` in the `MODELS` object.
+
+Sampling defaults are set in `onnx-streaming.js` (constructor):
+- `temperature`
+- `topK`
+- `topP`
+- `repetitionPenalty`
+
+---
+
+## Troubleshooting
+
+- **"Load failed" / model never becomes Ready**
+  - Verify the `onnx/` filenames match `MODELS` in `onnx-streaming.js`
+  - Check DevTools → Network for a missing `.onnx.data` file (404)
+  - Confirm `/` contains `tokenizer.json` (and related files)
+- **Performance notes**
+  - **Web Worker:** Keeps the UI responsive (no lag during generation).
+  - **int8 Decoder:** Optimized for high-throughput CPU inference.
+  - Achieves real-time streaming on modern hardware
+
+---
+
+## License & attribution
+
+Soprano is released under **Apache-2.0** in the upstream repository:
+https://github.com/ekwek1/soprano
+
+
+
+