QORA-TTS 1.7B - Pure Rust Text-to-Speech with Voice Cloning

Pure Rust TTS engine with voice cloning. No Python, no CUDA, no safetensors needed. Single executable + Q4 binary = portable TTS.

Based on Qwen3-TTS-12Hz-1.7B-Base (Apache 2.0).

Downlod 🤗: https://huggingface.co/qoranet/QORA-TTS/

Downlod 🤗: https://huggingface.co/qoranet/QORA-TTS-LIGHT

Quick Start

# Voice cloning with included voice
qora-tts.exe --model-path . --ref-audio voices/luna.wav --text "Hello, how are you?" --language english

# Different voice
qora-tts.exe --model-path . --ref-audio voices/adam.wav --text "Good morning!" --language english

# Clone your own voice (any 24kHz WAV)
qora-tts.exe --model-path . --ref-audio my_recording.wav --text "Custom voice" --language english

# Control length (codes = seconds x 12.5)
qora-tts.exe --model-path . --ref-audio voices/luna.wav --text "Short" --max-codes 100

# Custom output path
qora-tts.exe --model-path . --ref-audio voices/sagar.wav --text "Hi there" --language english --output greeting.wav

Files

  qora-tts.exe          4.3 MB   Inference engine
  model.qora-tts     1559 MB     Q4 weights (talker + predictor + decoder + speaker encoder)
  config.json           4.4 KB   Model configuration
  tokenizer.json         11 MB   Tokenizer (151,936 vocab)
  vocab.json            2.7 MB   Vocabulary
  merges.txt            1.6 MB   BPE merges
  tokenizer_config.json 7.2 KB   Tokenizer config
  voices/                         Reference WAV files for voice cloning

No safetensors needed. Everything loads from model.qora-tts.

Model Info

Property	Value
Base Model	Qwen3-TTS-12Hz-1.7B-Base
Type	Voice cloning only (no built-in speakers)
Quantization	Q4 (4-bit symmetric, group_size=32)
Binary Size	1559 MB
Sample Rate	24 kHz mono WAV
Languages	English, Chinese, German, Italian, Portuguese, Spanish, Japanese, Korean, French, Russian

Architecture

Component	Details
Talker	28 layers, hidden=2048, 16/8 GQA heads, SwiGLU 6144
Code Predictor	5 layers, hidden=1024, 16 code groups
Speech Decoder	8-layer transformer + Vocos vocoder
Speaker Encoder	ECAPA-TDNN (3 Res2Net blocks, 2048-dim output)

CLI Arguments

Flag	Default	Description
--model-path <dir>	.	Directory containing model.qora-tts + config
--text <text>	"Hello, how are you today?"	Text to synthesize
--ref-audio <wav>	-	Required - reference WAV for voice cloning
--language <name>	english	Target language
--output <path>	output.wav	Output WAV path
--max-codes <n>	500	Max code timesteps (~n/12.5 seconds)

Included Voices

Female	Male
luna, anushri, beth, caty, cherie, ember, faith, hope, jessica, kea, riya, vidhi, velvety	adam, charles, david, hale, heisenberg, joe, peter, quentin, sagar, steven, titan, true

All 24kHz WAV files in voices/. Use any 3-10 second clean speech recording for custom voice cloning.

Performance (i5-11500, 16GB RAM, CPU-only)

Phase	Time
Model Load	~0.8s (from binary)
Voice Extraction	~5-10s
Prefill	~3-8s
Code Generation	~2.5s/code
Audio Decode	~0.5s/frame
Memory	~1560 MB

Converting from Safetensors

If you have the original safetensors, convert to binary:

qora-tts.exe --model-path <safetensors_dir> --save model.qora-tts --text "x" --max-codes 1

This creates a single binary with all weights (talker + predictor + decoder + speaker encoder). After conversion, safetensors files are no longer needed.

---

Built with QORA - Pure Rust AI Inference