We Beat Whisper Large V3 On LibriSpeech With A 634 MB Model Running Entirely On Apple Silicon — Open Source Swift Library

Visit Tool

Shypd is an open-source Swift library that offers on-device speech AI, outperforming Whisper Large v3 on LibriSpeech. It provides speech-to-text, text-to-speech, and speech-to-speech capabilities, running entirely on Apple Silicon.

Claim this tool

No Views Yet

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is We beat Whisper Large v3 on LibriSpeech with a 634 MB model running entirely on Apple Silicon — open source Swift library?

speech-swift is an open-source Swift library designed for on-device speech AI on Apple Silicon. It provides advanced speech recognition (ASR) models like Qwen3-ASR and Parakeet TDT, which surpass OpenAI's Whisper Large v3 in accuracy and speed, running up to 40x real-time. The library also includes text-to-speech (TTS) with voice cloning, speech-to-speech, voice activity detection (VAD), speaker diarization, and speech enhancement. All functionalities operate locally on Apple devices, leveraging MLX for GPU inference and CoreML for the Neural Engine, ensuring privacy and efficiency without cloud dependencies or API calls.

Best used for

Ideal for developers and data scientists who need to integrate high-performance speech AI directly into Apple Silicon applications, process audio without cloud reliance, and build privacy-centric voice features. Especially valuable for creating fast, efficient, and accurate on-device speech solutions.

Common actions

transcribe audio

generate speech

process audio locally

implement speech AI

Apple Siliconaudio processingSwiftmachine learningopen-sourceon-device AISpeech Recognition

Capabilities

Key features

On-device speech recognition
Text-to-speech with voice cloning
Speech-to-speech conversion
Voice activity detection
Speaker diarization
Speech enhancement

Target Audience

developerdata scientist

Integrations

Not yet documented

Pricing & Plans

Open Source

FAQs

What Apple devices can run speech-swift models?

speech-swift models are optimized for Apple Silicon, meaning they can run on M-series Macs and iPhones. They leverage MLX for GPU inference and CoreML for the Neural Engine, ensuring efficient performance across compatible devices.

How does speech-swift compare to OpenAI's Whisper Large v3?

speech-swift outperforms Whisper Large v3 on the LibriSpeech benchmark, achieving better accuracy with significantly smaller model sizes. It also runs up to 40x faster than real-time on Apple Silicon, without requiring cloud access or API keys.

Does speech-swift support languages other than English?

Yes, Qwen3-ASR supports 52 languages, and Parakeet TDT supports 25 languages. While European languages perform well across both 4-bit and 8-bit quantization, 8-bit quantization is recommended for non-English languages to maintain accuracy.

Trending

Subcategories trending in Coding & Development

Code Assistants DevOps & Infrastructure No-Code / Low-Code Testing & QA Backend & APIs Prompt Engineering

Trending

Explore

Browse AI tools by category

Content & Design Productivity & Business Coding & Development AI Agents & Automation Research & Education Wellness & Lifestyle Career Development Marketing & Growth Data & Analytics Customer Support & CX Finance E-commerce