Whistle Launches 16.9 MB On-Device Speech-to-Text Model with Seven-Language Support and Word-Level Timestamps

1 reports archived1 independent sourcesupdated 10/9/2026, 00:59:39

Synthesis & Latest Updates

1 Sources Cross-Validated

Cactus Compute has released Whistle, a 16.9 MB ultra-lightweight on-device speech-to-text model built on the Needle architecture. It supports seven languages (English, German, French, Spanish, Italian, Dutch, Polish) and runs entirely offline in browsers or local devices, offering real-time transcription, word-level timestamps, and speech embeddings, significantly lowering the barrier for privacy-sensitive ASR deployment.

LATEST/Cactus Compute has released Whistle, a 16.9 MB ultra-lightweight on-device speech-to-text model built on the Needle architecture. It supports seven languages (English, German, French, Spanish, Italian, Dutch, Polish) and runs entirely offline in browsers or local devices, offering real-time transcription, word-level timestamps, and speech embeddings, significantly lowering the barrier for privacy-sensitive ASR deployment.

TIMELINECoverage timeline

Total 1 reports · Latest first

Hacker News AIT2·Oct 9, 00:5978 pts
Whistle Launches 16.9 MB On-Device Speech-to-Text Model with Seven-Language Support and Word-Level Timestamps
Original: Whistle: Speech to Text in 16.9 MB
- The model is only 16.9 MB and runs fully offline within browser tabs, ensuring audio data never leaves the device for privacy.
- Supports seven major European languages including English, German, and French with automatic language detection, handling up to 30 seconds of audio per pass.
- Beyond basic transcription, it provides word-level timestamps with probabilities derived from decoder attention and speech embeddings without requiring decoding.
View Evidence Original Source