Call Transcription

Every phone call made through Sarudo is automatically transcribed using faster-whisper, an optimized implementation of OpenAI's Whisper speech recognition model. The transcription runs entirely on your dedicated server — no audio data is sent to any external service. This means your call recordings and transcriptions are completely private and never leave your infrastructure. The transcription process starts immediately after a call ends and typically completes within seconds.

The faster-whisper model provides high-accuracy transcription for most business conversations. It handles multiple accents, industry terminology, and mixed-language conversations well. Transcription accuracy is typically above 95% for clear phone audio. Background noise, poor connections, or heavy accents may reduce accuracy slightly. The AI automatically adds punctuation and speaker identification to make transcripts easy to read.

After a call, the transcription is immediately available for review. Your AI employee can summarize the key points, extract action items, identify commitments made, and log the call in your CRM — all from the transcript. You can ask to see the full transcript at any time. Transcriptions are searchable, so you can later ask "what did Mike say about the project timeline?" and get the exact quote from the call.

Call recording and transcription are subject to local laws. In many jurisdictions, you must inform the other party that the call is being recorded. Your AI employee can be configured to announce recording at the start of each call. Check your local regulations regarding call recording consent requirements. Transcriptions are stored on your dedicated server and are subject to the same security and encryption standards as all other Sarudo data.