whatmodel.sh / stt / SpeechBrain Transformer + TransformerLM (LibriSpeech)
SpeechBrain Transformer + TransformerLM (LibriSpeech)
↗ share
+ contribute
SpeechBrain Transformer + TransformerLM (LibriSpeech) is an open-weights speech-to-text model from SpeechBrain, released March 2022. Licence: apache-2.0.
Catalogue identity notes
Comparison-only. Exact checkpoint speechbrain/asr-transformer-transformerlm-librispeech @bcf56f7 (repo created 2022-03-02, card release '24-03-22', Apache-2.0): transformer encoder, joint CTC/attention decoder and Transformer LM trained on LibriSpeech 960h. Needed so the SpeechBrain cards pool has a second model on LibriSpeech test-clean (speechbrain-asr-wav2vec2-librispeech) and a third on test-other. No local command verified.
Score within the Open ASR English short-form WER (March 2026) comparison set only.
$ how it compares
The speech-to-text models ranked either side of SpeechBrain Transformer + TransformerLM (LibriSpeech), out of 176 with a score, then more from SpeechBrain.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 148 |  ESPnet Transformer 18-layer + RNN LM | 42±4 | local | ||
| 149 |  ABR asr-19m-v2-en | 41±5 | local | ||
| 150 | SpeechBrain Transformer + TransformerLM (LibriSpeech) | 40±7 | comparison only | ||
| 151 | ESPnet Multi-blank Transducer 8421 | 40±4 | local | ||
| 152 | ESPnet Conformer RNN-T (no LM) | 40±4 | local | ||
| more from SpeechBrain | |||||
| 126 | SpeechBrain Conformer (LargeScaleASR / Loquacious) | 47±5 | local | ||
| 157 | SpeechBrain wav2vec2 CTC (LibriSpeech) | 35±4 | local | ||
see all speech-to-text models · find the ones that fit your machine
$ what it's good at
other tests2 not in a task · 79% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LibriSpeech test-other WER lab | 5.5% ✎ | ▁▁▁▂▅▃▅▄█▂▂▁ beats 17/54 | 59% |
| LibriSpeech test-clean WER lab | 2.3% ✎ | ▁·▁·▁▁▁▁▃▇█▆ beats 44/100 | 19% |
$ where to run it
No verified API or download recorded yet · add one
Official weights and documentation: huggingface.co/speechbrain ↗
$ how the score was worked out · SpeechBrain Hugging Face ASR… 79%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 SpeechBrain Hugging Face ASR model cards: published LibriSpeech/VoxPopuli WER for its own checkpoints · lab-reported | 2 | 79% |
| prior · pull toward the middle | 21% | |
| weighted average | = 40.4 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json