🧬 Darwin-TTS-1.7B-Cross

💬 Text to Speak

📝 Quick Examples

🧬 Darwin Enhancement

LLM FFN blended — more expressive, emotional speech

🎙️ Voice Cloning (Optional)

Upload 3~15 seconds of clear speech audio. The model will clone this voice.
If no audio is uploaded, a default voice will be used.

🎤

Click to upload or drag & drop
WAV, MP3, FLAC · 3~15 seconds recommended

🧬 Generating speech...
This may take 30~60 seconds on first run.

✨ Darwin-TTS Output

🔬 How It Works

Darwin-TTS v2 Architecture:
├── talker (28L Qwen3 LM) ← Layer-selective FFN blending (proprietary)
├── code_predictor (5L) ← untouched
├── speech_tokenizer ← untouched
└── encoder/decoder ← untouched

v2: Long text stability + quality exceeds original + Voice Cloning

Darwin-TTS-1.7B-Cross v2

🧬 Darwin Enhancement

🎙️ Voice Cloning (Optional)

✨ Darwin-TTS Output

🔬 How It Works