affectively-ai/whisper-small-onnx

Whisper Small (ONNX)

Production-ready ONNX conversion of openai/whisper-small for in-browser multilingual speech recognition — zero server cost, zero latency, complete privacy.

Highlights

• Multilingual ASR — supports 99 languages
• 244M parameters — lightweight enough for browser and mobile inference
• transformers.js compatible — drop-in pipeline('automatic-speech-recognition')
• Versatile — transcription, translation, and language detection

Quick Start

import { pipeline } from '@huggingface/transformers';

const transcriber = await pipeline(
  'automatic-speech-recognition',
  'affectively-ai/whisper-small-onnx'
);

const result = await transcriber(audioBlob);
// { text: 'Hello, how are you feeling today?' }

Model Details

Property	Value
Base model	openai/whisper-small
Parameters	244M
Languages	99
Input	16 kHz audio
Tasks	Transcription, translation, language ID
License	Apache 2.0

Use Cases

This model powers lightweight speech recognition in Edgework.ai — bringing fast, cheap, and private inference as close to the user as possible. Best for:

• Voice-to-text in multilingual emotion journaling
• Lightweight ASR where Parakeet is too large
• Language detection before routing to translation
• Quick voice notes and transcription

Related Models

Model	Parameters	Languages	Use case
whisper-small-onnx	244M	99	Lightweight multilingual
whisper-large-v3-onnx	1.5B	99+	Best multilingual quality
whisper-large-v3-turbo-onnx	809M	99+	Fast, high-quality
parakeet-ctc-0.6b-onnx	0.6B	English	Best English-only ASR

About

Published by AFFECTIVELY · Managed by @buley

We convert, quantize, and publish production-ready ONNX models for edge and in-browser inference. Every release is tested for correctness and stability before publication.

• All models · GitHub · Edgework.ai