Tether Launches QVAC TranscribePsy, a Local AI Speech-Recognition Model
- Tether AI Research introduced QVAC TranscribePsy, a multilingual speech-recognition model with 254 million parameters.
- The model can transcribe audio on-device, processing ten minutes of audio in just 1.65 seconds on an iPhone 17 Pro Max.
- It supports transcription in up to 25 European languages on laptops and eight languages on mobile devices.
- TranscribePsy achieved a word error rate (WER) of 8.02% on the Open ASR Leaderboard, indicating high accuracy.
- In performance tests, it ran 2.6 times faster than its closest competitor across six smartphones.
TranscribePsy is designed for developers seeking voice AI solutions that do not rely on cloud infrastructure, enhancing privacy for sensitive conversations like medical and legal discussions.
With its rapid transcription capabilities and local processing features, TranscribePsy represents a significant advancement in speech recognition technology, outperforming larger models while maintaining user privacy standards.
