Panels

Founded by Jason Le & Aaron Wenk in 2025

Provides audio datasets for training voice AI models.

Cost

Demo

Rating

★ Mixed Reviews

Time to value

Long Setup (> 1 day)

You can use Panels to get high-quality audio datasets for training voice AI models. The service provides different types of datasets including Clamor for speech-to-text training with noisy audio, Crisp for text-to-speech with clear audio, and Convo with real conversations between multiple speakers. They offer datasets in over 9 languages through their Rosetta collection and can create custom audio datasets tailored to your specific needs.

What Panels does

Download audio datasets for model trainingRequest samples to evaluate data qualitySpecify custom dataset requirements for your projectScale up datasets to match training needsCollect multi-channel conversation recordingsAnnotate audio files with transcriptionsProcess noisy audio for speech-to-text trainingGenerate clear audio samples for text-to-speech modelsProvides noisy single-channel audio for speech-to-text trainingOffers clear high-quality audio for text-to-speech modelsIncludes real conversations between multiple speakersSupports datasets in over 9 languagesCreates custom audio datasets based on specific requirementsCollects and annotates large-scale audio dataBuilds research-grade datasets for AI companiesProcesses audio through proprietary data pipelines