Loading
Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
MOSS-TTS-Nano (100M, CPU) fine-tunado em 4 vozes PT-BR (3M+1F, refs reais CML-TTS, 5k clips/voz via Qwen3-TTS-1.7B clone, 3 epocas). WER por voz: 05=0.10, 07=0.08, 11=0.05, 12=0.03 (Whisper-large-v3). Modo...
Model source
Source description
MOSS-TTS-Nano (100M, CPU) fine-tunado em 4 vozes PT-BR (3M+1F, refs reais CML-TTS, ~5k clips/voz via Qwen3-TTS-1.7B clone, 3 epocas). WER por voz: 05=0.10, 07=0.08, 11=0.05, 12=0.03 (Whisper-large-v3). Modo voice_clone: passe o audio de referencia da voz desejada.
Sources
1 sourceVerified Sep 19
Model artifacts
1 artifactSource excerpts
2 excerpts--- license: apache-2.0 language: - pt pipeline_tag: text-to-speech base_model: OpenMOSS-Team/MOSS-TTS-Nano tags: - tts - portuguese --- # moss-nano-ptbr-4voices MOSS-TTS-Nano (100M, CPU) fine-tunado em 4 vozes PT-BR (3M+1F, refs reais CML-TTS, ~5k clips/voz via Qwen3-TTS-1.7B clone, 3 epocas). **WER por voz: 05=0.10, 07=0.08, 11=0.05, 12=0.03** (Whisper-large-v3). Modo voice_clone: passe o audio de referencia da voz desejada.
Source context: 53 downloads · 0 likes · Pipeline text-to-speech · Repo marcosremar2/moss-nano-ptbr-4voices
