Refresh README — uniform WindyWord template with WER tier + dialect notes
Browse files
README.md
CHANGED
|
@@ -4,36 +4,57 @@ tags:
|
|
| 4 |
- automatic-speech-recognition
|
| 5 |
- whisper
|
| 6 |
- windyword
|
|
|
|
|
|
|
| 7 |
library_name: transformers
|
| 8 |
pipeline_tag: automatic-speech-recognition
|
| 9 |
language:
|
| 10 |
- en
|
|
|
|
| 11 |
---
|
| 12 |
|
| 13 |
# WindyWord.ai STT — Windy Lite
|
| 14 |
|
| 15 |
-
|
| 16 |
|
| 17 |
-
##
|
| 18 |
|
| 19 |
-
-
|
| 20 |
-
-
|
| 21 |
-
-
|
| 22 |
-
- `onnx-int8/`
|
| 23 |
|
| 24 |
-
##
|
| 25 |
|
| 26 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
|
| 28 |
-
|
| 29 |
|
| 30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
-
##
|
| 35 |
|
| 36 |
-
|
| 37 |
|
| 38 |
---
|
| 39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
- automatic-speech-recognition
|
| 5 |
- whisper
|
| 6 |
- windyword
|
| 7 |
+
- english
|
| 8 |
+
- multilingual
|
| 9 |
library_name: transformers
|
| 10 |
pipeline_tag: automatic-speech-recognition
|
| 11 |
language:
|
| 12 |
- en
|
| 13 |
+
- multilingual
|
| 14 |
---
|
| 15 |
|
| 16 |
# WindyWord.ai STT — Windy Lite
|
| 17 |
|
| 18 |
+
**Multilingual speech-to-text engine. Transcribes audio in 100+ languages, with English as the primary trained domain.**
|
| 19 |
|
| 20 |
+
## Profile
|
| 21 |
|
| 22 |
+
- **Architecture:** 74M params · whisper-base
|
| 23 |
+
- **Profile:** fast
|
| 24 |
+
- **Base model:** [openai/whisper-base](https://huggingface.co/openai/whisper-base)
|
|
|
|
| 25 |
|
| 26 |
+
## Variants in this repo
|
| 27 |
|
| 28 |
+
| Subfolder | Format | Use case |
|
| 29 |
+
|---|---|---|
|
| 30 |
+
| `safetensors/` | PyTorch safetensors (FP32) | GPU inference (highest quality) |
|
| 31 |
+
| `ct2-int8/` | CTranslate2 INT8 | CPU inference (~25% size, 2-4× faster) |
|
| 32 |
+
| `onnx/` | ONNX FP32 | Cross-platform deployment |
|
| 33 |
+
| `onnx-int8/` | ONNX INT8 | Edge / mobile / WebAssembly |
|
| 34 |
|
| 35 |
+
## Usage
|
| 36 |
|
| 37 |
+
```python
|
| 38 |
+
from transformers import WhisperForConditionalGeneration, WhisperProcessor
|
| 39 |
+
processor = WhisperProcessor.from_pretrained("WindyWord/listen-windy-lite", subfolder="safetensors")
|
| 40 |
+
model = WhisperForConditionalGeneration.from_pretrained("WindyWord/listen-windy-lite", subfolder="safetensors")
|
| 41 |
+
```
|
| 42 |
|
| 43 |
+
For CPU inference via CTranslate2:
|
| 44 |
+
```python
|
| 45 |
+
import ctranslate2
|
| 46 |
+
# After downloading the ct2-int8 subfolder:
|
| 47 |
+
model = ctranslate2.models.Whisper("path/to/ct2-int8/")
|
| 48 |
+
```
|
| 49 |
|
| 50 |
+
## Commercial Use
|
| 51 |
|
| 52 |
+
Part of the [WindyWord.ai](https://windyword.ai) STT fleet. Visit windyword.ai for real-time voice-to-text + translation apps and API access.
|
| 53 |
|
| 54 |
---
|
| 55 |
+
|
| 56 |
+
## Provenance & License
|
| 57 |
+
|
| 58 |
+
Weights derived from [openai/whisper-base](https://huggingface.co/openai/whisper-base) under Apache-2.0 (inherited). Proprietary fine-tuning by WindyWord.ai team via LoRA fog-of-mirror methodology where applicable.
|
| 59 |
+
|
| 60 |
+
*Certified by Opus 4.6 Opus-Claw (Dr. C) on Veron-1 (RTX 5090, Mt Pleasant SC).*
|