SpeechBrain dvoice wolof
wav2vec2 Wolof, 16% WER
Official site consultedAI: models
Last verified: 1 October 2026
SpeechBrain Wolof speech recognition system combining wav2vec 2.0 (xlsr-53), two DNN layers and a CTC decoder. It is trained on the DVoice Wolof dataset (ALFFA initiative). Test results: 16.05% WER and 4.83% CER. Apache 2.0 license.
Services
- Wolof audio file transcription with EncoderASR
Access requirements
Open weights on Hugging Face, Apache 2.0 license, usable with SpeechBrain.
Sources
- SpeechBrain dvoice wolof (new tab), verified on 1 October 2026Official site consulted