AI news ·
Salesforce Releases Open-Weight SOV-Transcribe-1 Speech Model
Salesforce released an open-weight speech-to-text model that scores well on accuracy tests across multiple English accents and business audio.
Salesforce has released SOV-Transcribe-1, a 4-billion-parameter speech-to-text model, under the Apache 2.0 license. The company reports that the open-weight model achieves leading results on speech recognition benchmarks, including tests across 14 English accents and earnings call audio.
The model is available for teams working in customer support, sales, education, marketing, product development, and other fields where accurate transcription matters. Users should run their own evaluations on accuracy, bias, and latency using their specific audio data before adopting the model in production.