AI Tools.

Search

wav2vec2-base

Wav2Vec2-base is Meta's self-supervised speech model pretrained on 960 hours of LibriSpeech, producing 768-dimensional audio representations. Fine-tuning on labeled data produces competitive ASR results with limited labeled data.

Summary text generated by an automated pipeline from the model card · Not individually reviewed or run by us · How this page is made

From the model card

Fields below are copied from the tags and counters on the HuggingFace repository facebook/wav2vec2-base at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.

Publisher (HF namespace)
facebook
Library
Transformers
Framework tags
PyTorch
License tag
apache-2.0 — read the license file in the repo before relying on it
Language tags
English (en)
Papers cited
arXiv:2006.11477
Datasets declared
librispeech_asr
Downloads (HF counter at last fetch)
3,115,743
Likes (HF counter at last fetch)
124
Model card
https://huggingface.co/facebook/wav2vec2-base

Use cases

  • ASR fine-tuning on domain-specific speech data
  • Speech representation learning for downstream classification
  • Low-resource language ASR when labeled data is scarce
  • Academic baseline for self-supervised speech model research

Pros

  • Strong ASR fine-tuning results with as few as 10 minutes of labeled data
  • MIT licensed
  • Standard baseline in self-supervised speech learning literature
  • Available in base and large variants for scaling experiments

Cons

  • Superseded by wav2vec2-large and Whisper for most practical ASR applications
  • English-focused pretraining — limited multilingual transfer
  • Requires fine-tuning on labeled data — not zero-shot like Whisper
  • No built-in language identification or speaker information

Tags

transformerspytorchwav2vec2pretrainingspeechendataset:librispeech_asrarxiv:2006.11477license:apache-2.0endpoints_compatibleregion:us