Modelpublic

LocalAI-io/vibevoice-asr-it-yodas-only-ggml

sourceHugging Facemitupdated 5mo agoView on Hugging Face

0likes101downloads

Model Card

vibevoice-asr-it-yodas-only-ggml

GGUF quantizations for vibevoice.cpp of LocalAI-io/vibevoice-asr-it-yodas-only — VibeVoice-ASR fine-tuned on YODAS-only Italian (asr_only ~87k + ast 200k cap, conversational YouTube audio).

Files

File	Quantization	Size	Notes
`vibevoice-asr-it-yodas-only-q8_0.gguf`	Q8_0	~13 GB	Near-lossless, recommended for accuracy
`vibevoice-asr-it-yodas-only-q5_0.gguf`	Q5_0	~11 GB	Sweet spot quality/size
`vibevoice-asr-it-yodas-only-q4_0.gguf`	Q4_0	~10 GB	Smallest, fastest

Usage

bash

hf download LocalAI-io/vibevoice-asr-it-yodas-only-ggml vibevoice-asr-it-yodas-only-q5_0.gguf --local-dir .
./build/bin/vibevoice-cli asr --model vibevoice-asr-it-yodas-only-q5_0.gguf --audio audio.wav

See vibevoice.cpp for build + tokenizer setup.