Audio data platform that designs and delivers high-quality multilingual speech datasets for training and evaluating voice AI models.
Panels produces custom and off-the-shelf audio datasets for voice AI labs and startups, covering high-quality speaker-separated audio, single-speaker scripted recordings, and turn-taking evaluation datasets across multiple languages. The company handles end-to-end data collection, quality assurance, transcription, and review in-house, working with clients to define target use cases, speakers, environments, and edge cases with clear acceptance criteria. Datasets can be purchased with exclusivity rights, and samples for bespoke projects are typically delivered within two weeks. Panels works with both early-stage startups and large frontier labs. Panels is a product of Panels.