Listen integrates two leading transcription engines - OpenAI's Whisper and Gladia - and lets you choose between them in Settings. Both are excellent; neither is universally superior. The choice depends on your content and your priorities.
Whisper: the reliable workhorse
Whisper is Listen's default engine. It performs consistently across standard British and American English, as well as French, Spanish and a broad range of other languages. For most professional use cases - team meetings, structured interviews, lectures - it delivers clean transcripts with minimal intervention. It processes quickly, consumes one unit of quota per minute recorded, and integrates seamlessly with Listen's AI summary and action-item extraction.
Gladia: precision for accents and multilingual content
Gladia was built with accent diversity and multi-speaker scenarios as primary design goals. If you regularly record sources with strong regional or non-native accents, or conduct international calls where participants switch between languages, Gladia will outperform Whisper on accuracy. The trade-off: it consumes twice the quota (one minute recorded = two minutes of quota). See our comparison: international multilingual meetings. Note that AI summaries and decision/task extraction use GPT-4o separately - available with both engines regardless of which transcription option you choose.
If a Whisper transcription disappoints, use the Re-transcribe button in the recording detail view, switch to Gladia in Settings, and the original audio will be reprocessed - no need to re-record.
Compare our plans to find the right quota for your Whisper/Gladia mix.