Q
Qwen3-TTS Base 1.7B
bởi QwenCông khaiapache-2.0
README.md
Qwen3-TTS Base 1.7B
Qwen3-TTS Base 1.7B is a high-fidelity zero-shot voice cloning text-to-speech model from the Alibaba Qwen team, published under the Qwen organization on Hugging Face.
Links
- Upstream model card: https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-Base
- Upstream GitHub: https://github.com/QwenLM/Qwen2.5-TTS
Model Facts
- Task: Text-to-Speech (Zero-Shot Voice Cloning)
- Parameters: 1.7B
- License: Apache-2.0
- Architecture: Qwen2.5-TTS GGUF
LA Studio Notes
Qwen3-TTS Base 1.7B in LA Studio is a zero-shot voice cloning workflow. It downloads GGUF models from conversion repository cstr/qwen3-tts-1.7b-base-GGUF. It needs:
- Backbone:
qwen3-tts-12hz-1.7b-base-q8_0.gguforqwen3-tts-12hz-1.7b-base-f16.gguf - Tokenizer:
qwen3-tts-tokenizer-12hz.gguf
It requires a 24kHz reference WAV file and a reference transcript to clone a voice.
Cấu hình tham số quy trình
Cấu hình các tham số tùy chỉnh này để tinh chỉnh quy trình tiếng nói và giọng nói.
Temperaturefloat
Controls randomness of speech generation. Higher values increase variety.
1
Seedint
Random seed for reproducibility. Set to -1 for random.
-1
Max Codec Stepsint
Maximum decoding steps for the neural voice audio synthesizer.
100