Frequently Asked Questions

Lucent is a podcast transcript search tool. You search for any podcast by name, pick the episodes you want to analyse, and Lucent transcribes them using OpenAI Whisper. Once transcribed, you can search any word or phrase across every episode and get timestamped excerpts showing exactly where that topic was discussed.

Any podcast with a public RSS feed works. That covers the vast majority of shows — independent podcasts, major network shows, and everything in between. Lucent uses the podcast's public feed to fetch audio files, so password-protected or private feeds are not currently supported.

Lucent uses Whisper, which achieves roughly 4–7% word error rate on clear English speech — better than most human transcribers for fast speakers. Accuracy is slightly lower for heavy accents, technical jargon, or episodes with significant background noise. Timestamps are generated at the word level by Whisper and are accurate to within a second.

Transcription runs at roughly 5–10× real time. A one-hour episode typically finishes in 6–12 minutes. You can add more episodes while earlier ones are still processing — Lucent queues them automatically. Once an episode is transcribed, it is cached, so adding it to a second collection is instant.

A collection groups episodes from one or more podcasts around a research topic. You give the collection a name (e.g., "AI regulation coverage"), select the episodes you care about, and Lucent keeps them together so you can search across the whole set. You can have as many collections as you like.

Yes — that is the core feature. Type a topic or phrase, and Lucent returns every mention across all episodes in the collection, with a timestamp for each. It also generates a cross-episode recap that synthesises what was said on that topic across the entire set.

Yes. Each collection has a Share button that generates a read-only link. Anyone with the link can search and browse the transcripts — no account required. The link owner can revoke access at any time. The owner dashboard also shows who opened the link and which topics visitors searched.

No. Transcripts are stored once and reused. If an episode you pick has already been transcribed (by you or anyone else on the platform), it is added instantly at no extra processing cost. This makes building focused sub-collections very fast.

No. Search for the podcast by name. Lucent fetches the full episode list from the RSS feed — you just check the episodes you want. No URLs, no file downloads on your end.

During the current period there are no hard caps. Very large collections (100+ episodes) take longer to transcribe but work correctly. If you are planning a large research project, reaching out to let us know helps us prioritise capacity.