Metadata-Version: 2.4
Name: f5-myanmar-tts
Version: 0.2.1
Summary: Official Python API and CLI for F5-TTS Burmese (794.5h Pretrained Foundation Model)
Author: freococo
License: CC-BY-NC-4.0
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Multimedia :: Sound/Audio :: Speech
Classifier: Operating System :: OS Independent
Requires-Python: >=3.9
Description-Content-Type: text/markdown
Requires-Dist: torch
Requires-Dist: torchaudio
Requires-Dist: f5-tts
Requires-Dist: vocos
Requires-Dist: huggingface_hub
Requires-Dist: soundfile

# f5-myanmar-tts 🇲🇲 (v2)
Official Python library and CLI for the **F5-TTS Burmese v2 Foundation Model** (**1,025,604 updates / 36 epochs**).

Now featuring **50% smaller download size** (~643 MB FP16 model) and native Burmese syllable-aware pacing.

Automatically downloads and caches model weights directly from Hugging Face: `freococo/F5-Myanmar-TTS`.

---

## ⚡ Quickstart (CLI)

```bash
# Speak any Burmese sentence with default voice
f5-myanmar "လူတွေ အားလုံးကို ချစ်မြတ်နိုးပါ တန်ဖိုးထားပါ လေးစားစွာ ဆက်ဆံပါ" -o output.wav

# Clone your own voice (3 to 5 seconds of audio)
f5-myanmar "မင်္ဂလာပါ ခင်ဗျာ" --ref-audio my_voice.wav --ref-text "ကျွန်တော် မောင်မောင်ပါ" -o custom.wav
```

---

## 🐍 Python API

```python
from f5_myanmar_tts import MyanmarTTS

# Auto-downloads lightweight v2 FP16 model (~643MB) on first run
tts = MyanmarTTS()

# Generate Burmese speech
tts.speak(
    "လူတွေ အားလုံးကို ချစ်မြတ်နိုးပါ တန်ဖိုးထားပါ လေးစားစွာ ဆက်ဆံပါ", 
    output_file="speech.wav"
)
```

---

## 🌟 What's New in v2
- **1,025,604 Updates:** Trained past 1 Million updates (36 epochs) for superior tone and syllable consistency.
- **FP16 Optimized:** Pruned model weights down from 1.3 GB to **643 MB** for 2x faster downloads.
- **Natural Pacing:** Automatic silence trimming and duration calculations to prevent cut-offs.

## 📜 License & Dedication
Released under **CC BY-NC 4.0** for free research, cultural preservation, education, and open-source innovation.
