{"model_name": "LeanVoice", "display_name": "[LeanVoice ](https://leanvoice.ai) [\ud83e\udd17](https://huggingface.co/spaces/zyushg/leanvoice-tts)", "model_tags": ["Non-autoregressive", "Diffusion"], "web_url": "https://leanvoice.ai", "hf_url": "https://huggingface.co/spaces/zyushg/leanvoice-tts", "code_url": "", "paper_url": "", "inference_details": "Closed-source proprietary model served via api.leanvoice.ai. Fixed-voice system (6 voices) \u2014 does NOT do zero-shot speaker cloning, so prompts were synthesized in the model's own voices (rotated across the set for speaker diversity) rather than cloning the reference speakers. Real-time CPU inference, 8 steps, speed 1.05, 44.1kHz mono WAV.", "status": "pending"}