Simplify PnC quickstart display
Browse files
README.md
CHANGED
|
@@ -39,16 +39,18 @@ is a separate speech recognizer; PnC v2 can process its output.
|
|
| 39 |
|
| 40 |
## Quickstart
|
| 41 |
|
| 42 |
-
|
|
|
|
| 43 |
|
| 44 |
```bash
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
# Hej, mit navn er Emil. Hvordan går det i dag?
|
| 48 |
```
|
| 49 |
|
| 50 |
-
|
| 51 |
-
|
|
|
|
|
|
|
| 52 |
For longer transcripts, the runtime uses 128-token windows with 12 words of overlap
|
| 53 |
and keeps every input word in order. Pass continuous text when possible so PnC can
|
| 54 |
use context across speech recognition chunks.
|
|
|
|
| 39 |
|
| 40 |
## Quickstart
|
| 41 |
|
| 42 |
+
Clone the [Ekko STT repository](https://github.com/Rye-A1/ekko-danish-stt),
|
| 43 |
+
then run this from its root with [uv](https://docs.astral.sh/uv/) and Python 3.11+:
|
| 44 |
|
| 45 |
```bash
|
| 46 |
+
uv sync --locked --no-editable --extra pnc
|
| 47 |
+
uv run --no-sync ekko pnc "hej mit navn er emil hvordan går det i dag"
|
|
|
|
| 48 |
```
|
| 49 |
|
| 50 |
+
Expected output: `Hej, mit navn er Emil. Hvordan går det i dag?`
|
| 51 |
+
|
| 52 |
+
The first command installs the inference-only runtime. The first PnC run
|
| 53 |
+
downloads the pinned int8 model from Hugging Face; later runs use the cache.
|
| 54 |
For longer transcripts, the runtime uses 128-token windows with 12 words of overlap
|
| 55 |
and keeps every input word in order. Pass continuous text when possible so PnC can
|
| 56 |
use context across speech recognition chunks.
|