Automatic Speech Recognition
NeMo
ONNX
GGUF
Danish
speech-recognition
speech-to-text
danish
danish-asr
offline
local-inference
cpu-inference
sherpa-onnx
rnnt
word-timestamps
Eval Results (legacy)
Instructions to use RyeAI/ekko-v1-tiny with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- NeMo
How to use RyeAI/ekko-v1-tiny with NeMo:
import nemo.collections.asr as nemo_asr asr_model = nemo_asr.models.ASRModel.from_pretrained("RyeAI/ekko-v1-tiny") transcriptions = asr_model.transcribe(["file.wav"]) - Notebooks
- Google Colab
- Kaggle
Describe clear-speech and dictation use
Browse files
README.md
CHANGED
|
@@ -98,6 +98,11 @@ word timestamps. It is a FastConformer RNN-T fine-tuned from
|
|
| 98 |
`nvidia/parakeet-rnnt-110m-da-dk` and runs locally. Its raw output is lowercase
|
| 99 |
text; pair it with `RyeAI/ekko-pnc` for capitalization and punctuation.
|
| 100 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 101 |
## Benchmark
|
| 102 |
|
| 103 |
**12.37% mean WER** across 26,780 rows from five Danish test sets, evaluated
|
|
@@ -214,7 +219,8 @@ pseudo-labelled audio.
|
|
| 214 |
- The 44-token vocabulary does not contain uppercase or punctuation tokens.
|
| 215 |
- Punctuation and capitalization require a separate restoration model such as
|
| 216 |
`RyeAI/ekko-pnc`.
|
| 217 |
-
- Spontaneous conversation remains the hardest benchmark domain.
|
|
|
|
| 218 |
|
| 219 |
## License
|
| 220 |
|
|
|
|
| 98 |
`nvidia/parakeet-rnnt-110m-da-dk` and runs locally. Its raw output is lowercase
|
| 99 |
text; pair it with `RyeAI/ekko-pnc` for capitalization and punctuation.
|
| 100 |
|
| 101 |
+
**Best fit:** clear Danish speech from one speaker, delivered at a steady pace,
|
| 102 |
+
such as dictation or reading aloud. The published results are stronger on read
|
| 103 |
+
and formal speech than on spontaneous conversation. Dictation itself was not
|
| 104 |
+
evaluated as a separate test set.
|
| 105 |
+
|
| 106 |
## Benchmark
|
| 107 |
|
| 108 |
**12.37% mean WER** across 26,780 rows from five Danish test sets, evaluated
|
|
|
|
| 219 |
- The 44-token vocabulary does not contain uppercase or punctuation tokens.
|
| 220 |
- Punctuation and capitalization require a separate restoration model such as
|
| 221 |
`RyeAI/ekko-pnc`.
|
| 222 |
+
- Spontaneous conversation remains the hardest benchmark domain (24.15% WER).
|
| 223 |
+
- Noisy audio and overlapping speakers were not evaluated separately.
|
| 224 |
|
| 225 |
## License
|
| 226 |
|