Project description
Our company accumulates a large volume of audio and video recordings every week: client calls, meetings, interviews, webinars, and podcast episodes. Currently all this material sits as dead weight — no one has time to listen to hours of recordings, so agreements and important quotes are lost immediately after the conversation. We need an automatic speech-to-text transcription service that turns each recording into an accurate and searchable transcript.
The foundation of the project is a reliable speech recognition engine. The service should accept common audio and video formats, transcribe them with high accuracy even on imperfect recordings, and add punctuation and paragraphs so the text reads normally. Diarization is mandatory: each line is attributed to a speaker, and word-level timestamps allow instant navigation to the right place in the recording. Batch processing is also needed — upload dozens of files at once.
No less important is what happens after recognition. We need a convenient editor where we can correct errors, rename speakers, merge and split segments, and search the entire archive by keywords. Ready transcripts should be exported to text, subtitles, and structured formats. Please describe your experience with speech recognition models, expected accuracy on real audio, supported languages, and how you'll ensure recording confidentiality.
What matters to us:
— automatic transcription with punctuation and speaker diarization;
— word-level timestamps and full-text search across the archive;
— editor for corrections and export to text, subtitles, and structured data.
The foundation of the project is a reliable speech recognition engine. The service should accept common audio and video formats, transcribe them with high accuracy even on imperfect recordings, and add punctuation and paragraphs so the text reads normally. Diarization is mandatory: each line is attributed to a speaker, and word-level timestamps allow instant navigation to the right place in the recording. Batch processing is also needed — upload dozens of files at once.
No less important is what happens after recognition. We need a convenient editor where we can correct errors, rename speakers, merge and split segments, and search the entire archive by keywords. Ready transcripts should be exported to text, subtitles, and structured formats. Please describe your experience with speech recognition models, expected accuracy on real audio, supported languages, and how you'll ensure recording confidentiality.
What matters to us:
— automatic transcription with punctuation and speaker diarization;
— word-level timestamps and full-text search across the archive;
— editor for corrections and export to text, subtitles, and structured data.
Other / Miscellaneous Cash, Bank transfer
