Features
Channel-separated natural two-speaker conversations (Converse dataset)
Multilingual dataset spanning 15+ languages with dialect/accent metadata (Atlas)
Multi-speaker conversations for speaker separation and diarization (Chorus)
Expert domain-specific conversations (Dialog dataset)
Rigorous six-step dataset development process (Hypothesize to Release)
Custom dataset design in partnership with research teams
High-quality audio data for speech-to-speech systems
Data suitable for training transcription, translation, and synthesis models
Continuous dataset improvement after publication
Rapid access: off-the-shelf datasets available within one to two days
Sample requests to evaluate before purchase
Data license agreements tailored to use case
Funding: $50M Series B in Oct 2025, $25M Series A in May 2025
Automatic speech-to-text transcription
Interactive transcript editor with timestamps
Speaker identification (diarization)
Export to SRT, VTT, TXT, DOCX
Human proofreading services
Subtitle creation and syncing
Media player with text highlighting
Team collaboration features
