stt_en_conformer_ctc_large
NVIDIA🇺🇸 USMajor lab·
Audioopen weights
The model transcribes English speech into lowercase letters, including spaces and apostrophes. It is a non-autoregressive Conformer with CTC decoding, usable for inference, fine-tuning, and deployment with NVIDIA Riva.
Summary of the model card
- License
- cc-by-4.0
- Modalities
- text → text
2 variants
Sources
HuggingFaceDeutsch
Das Modell transkribiert englische Sprache in Kleinbuchstaben inklusive Leerzeichen und Apostrophen. Es ist eine nicht autoregressive Conformer-Variante mit CTC-Decoding, die für Inferenz und Fine-Tuning genutzt und mit NVIDIA Riva ausgeliefert werden kann.
Benchmarks
| Benchmark | Domain | Value | Evidence |
|---|---|---|---|
| Earnings-22Open ASR Leaderboard (Langform) | audio-stt | 23.15 % WER | 68 |
| Earnings-22Open ASR Leaderboard (Langform) | audio-stt | 17.51 % WER | 68 |
| Open ASR LeaderboardOpen ASR Leaderboard (Langform) | audio-stt | 18.7275 % WER | 58 |
| Open ASR LeaderboardOpen ASR Leaderboard (Langform) | audio-stt | 14.4375 % WER | 58 |
| TED-LIUM 3Open ASR Leaderboard (Langform) | audio-stt | 6.26 % WER | 49 |
| TED-LIUM 3Open ASR Leaderboard (Langform) | audio-stt | 3.68 % 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
- —