README.md

---
license: apache-2.0
tags:
- automatic-speech-recognition
- whisper
- windyword
- english
- multilingual
library_name: transformers
pipeline_tag: automatic-speech-recognition
language:
- en
- multilingual
---

# WindyWord.ai STT — Windy Pro Engine

**Multilingual speech-to-text engine. Transcribes audio in 100+ languages, with English as the primary trained domain.**

## Profile

- **Architecture:** 1.55B params · whisper-large-v3
- **Profile:** premium / max accuracy
- **Base model:** [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3)

## Variants in this repo

| Subfolder | Format | Use case |
|---|---|---|
| `safetensors/` | PyTorch safetensors (FP32) | GPU inference (highest quality) |
| `ct2-int8/` | CTranslate2 INT8 | CPU inference (~25% size, 2-4× faster) |
| `onnx/` | ONNX FP32 | Cross-platform deployment |
| `onnx-int8/` | ONNX INT8 | Edge / mobile / WebAssembly |

## Usage

```python
from transformers import WhisperForConditionalGeneration, WhisperProcessor
processor = WhisperProcessor.from_pretrained("WindyWord/listen-windy-pro-engine", subfolder="safetensors")
model = WhisperForConditionalGeneration.from_pretrained("WindyWord/listen-windy-pro-engine", subfolder="safetensors")
```

For CPU inference via CTranslate2:
```python
import ctranslate2
# After downloading the ct2-int8 subfolder:
model = ctranslate2.models.Whisper("path/to/ct2-int8/")
```

## Commercial Use

Part of the [WindyWord.ai](https://windyword.ai) STT fleet. Visit windyword.ai for real-time voice-to-text + translation apps and API access.

---

## Provenance & License

Weights derived from [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) under Apache-2.0 (inherited). Voice tiers are direct redistributions of the upstream community Whisper / distil-whisper variants; no LoRA fine-tuning has been applied to these voice models.

*Certified by Opus 4.6 Opus-Claw (Dr. C) on Veron-1 (RTX 5090, Mt Pleasant SC).*