May 20
Add support Soniox v4 Real-Time STT Model
I’ve been heavily using Superwhisper for the past ~8 months and spent a lot of time comparing different STT models for real everyday use. After testing many of the currently available models, Soniox v4 Real-Time consistently stands out as the best experience I’ve had so far, especially for multilingual dictation and mixed-language speech.
What makes it stand out for me is the combination of speed, accuracy, and real-time responsiveness. The transcription appears almost instantly while speaking, and the overall latency feels noticeably lower than most other models I’ve tested.
It also handles multilingual speech surprisingly well. Most STT models still struggle when switching languages mid-sentence, mixing English technical terms into another language, speaking naturally and quickly, or using imperfect microphones/background noise. Soniox is one of the very few models I’ve tried that consistently handles these situations well in real time.
It also seems much better at maintaining punctuation, sentence continuity, technical terms, numbers, and more natural endpointing during dictation.
They also provide an open comparison tool where anyone can test the same audio against multiple providers side by side: https://soniox.com/compare
Also, I recently found out that even Perplexity uses Soniox for voice chat.
Would genuinely love to see this model available in Superwhisper.
Pending