Modelpublic

HamzaBrh/qwen3-tts-0.6b-bf16

sourceHugging Facemitupdated 6mo agoView on Hugging Face

0likes

Model Card

Qwen3-TTS 0.6b — 0.6b-bf16

Full precision bf16 (baseline)

Quantization Config

Parameter	Value
Backbone	none
Code Predictor	none
Vocoder CPU	False
Text Embed Offload	False

Performance

Metric	Value
VRAM	2675 MB
RTF	0.13 (7.7x real-time)
Weight Storage	2009 MB

Quality (Seed-TTS EN)

Metric	Value
WER (EN)	3.25%

Based On

Qwen3-TTS-12Hz-0.6B-Base

Built with MegaQwen TTS