stt_en_conformer_transducer_large
NVIDIA🇺🇸 USMajor lab·
Audioopen weights
Transcribes English speech into lowercase text and is a Conformer-Transducer model for NeMo. On LibriSpeech test-clean, version 1.10.0 reaches 1.7 percent WER with greedy decoding according to the model card.
Summary of the model card
- License
- cc-by-4.0
- Modalities
- text → text
2 variants
Sources
HuggingFaceDeutsch
Transkribiert englische Sprache in Kleinbuchstaben und ist ein Conformer-Transducer-Modell fuer NeMo. Auf LibriSpeech test-clean erreicht die Version 1.10.0 laut Modelcard 1.7 Prozent WER bei Greedy-Decoding.
Benchmarks
| Benchmark | Domain | Value | Evidence |
|---|---|---|---|
| Earnings-22Open ASR Leaderboard (Langform) | audio-stt | 23.69 % WER | 68 |
| Earnings-22Open ASR Leaderboard (Langform) | audio-stt | 20.72 % WER | 68 |
| Open ASR LeaderboardOpen ASR Leaderboard (Langform) | audio-stt | 21.9225 % WER | 58 |
| Open ASR LeaderboardOpen ASR Leaderboard (Langform) | audio-stt | 13.74333333 % WER | 58 |
| TED-LIUM 3Open ASR Leaderboard (Langform) | audio-stt | 5.46 % WER | 49 |
| TED-LIUM 3Open ASR Leaderboard (Langform) | audio-stt | 5.24 % WER | 49 |
Vendor-reported numbers are marked and count with factor 0.7 in rankings. The evidence score says how much a benchmark still tells you today.
Provenance
- Detected
- Aug 25, 2026
- First source
- openasr
- Description
- Summary of the model card
- Editorially reviewed
- —