AI Tools.

Search

audio classification by alefiury

wav2vec2-large-xlsr-53-gender-recognition-librispeech

wav2vec2-large-xlsr-53-gender-recognition-librispeech is a binary audio classifier fine-tuned from Facebook's wav2vec2-xls-r-300m to predict speaker gender. It was trained on LibriSpeech ASR data and applies the XLSR cross-lingual speech representation backbone to a classification head.

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 alefiury/wav2vec2-large-xlsr-53-gender-recognition-librispeech 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)
alefiury
Pipeline tag
audio-classification
Library
Transformers
Framework tags
PyTorch
Weight formats
safetensors
License tag
apache-2.0 — read the license file in the repo before relying on it
Lineage
Datasets declared
librispeech_asr
Downloads (HF counter at last fetch)
1,052,869
Likes (HF counter at last fetch)
48
Model card
https://huggingface.co/alefiury/wav2vec2-large-xlsr-53-gender-recognition-librispeech

Use cases

  • Speaker gender detection in call center audio pipelines
  • Demographic metadata extraction from podcast or broadcast audio
  • Audio dataset stratification by predicted speaker gender
  • Pre-screening step in speaker diarization workflows
  • Research into fairness and bias in speech classification systems

Pros

  • Fine-tuned from a 300M-parameter cross-lingual backbone, providing broad phoneme coverage
  • Apache 2.0 license allows unrestricted commercial integration
  • Narrow binary classification task yields fast inference with minimal compute
  • HuggingFace transformers-native, requires no custom inference code
  • LibriSpeech training data is well-documented, enabling reproducible evaluation

Cons

  • Trained exclusively on LibriSpeech (read English speech), limiting robustness on spontaneous or accented audio
  • Binary gender framing does not accommodate non-binary speaker demographics
  • XLS-R-300M backbone is large relative to the simplicity of the classification task
  • No published eval metrics on out-of-domain datasets in the model card
  • Predictions can reflect training data biases rather than acoustic gender markers

Tags

transformerspytorchsafetensorswav2vec2audio-classificationgenerated_from_trainerdataset:librispeech_asrbase_model:facebook/wav2vec2-xls-r-300mbase_model:finetune:facebook/wav2vec2-xls-r-300mlicense:apache-2.0endpoints_compatibleregion:us