Voice AI turns a recording into text and then into usable output. ## Before you upload - Record as cleanly as you can: one conversation, not a room full of background noise. - Check the file size against your plan's limit before starting. - Confirm you have permission to record the people in it. ## Choose a mode There are nine modes, and they are instructions for how the recording should be read rather than different models: General, Business, Meeting, Interview, Lecture & Learning, Voice Notes, Podcast & Content, Customer Call and E-commerce. Pick the one closest to the recording. **Meeting** keeps decisions, owners and deadlines exact; **Customer Call** labels customer and agent where that is clear; **E-commerce** keeps SKUs, prices and quantities exact. ## Verbatim or clean read **Verbatim** keeps the words as spoken, including filler. **Clean read** removes the stumbles and repeated words while keeping the meaning. Use verbatim for records and clean read for anything a person will read. ## Insights Once you have a transcript, ask for a summary, minutes, action items, a report or a content draft. Insights are generated from the transcript, so the quality of the transcript sets the ceiling. ## Allowances Voice AI is available on the Business and Enterprise plans. Business allows files up to 15 MB and 100 transcriptions a day; Enterprise allows 15 MB and 300 a day. A safe default of 10 MB and 20 a day applies where a plan cannot be resolved. ## Honest limits Speaker labels are AI estimates: where the audio is unclear, a change of speaker may be missed or mislabelled. Accuracy depends on audio quality, accents and background noise. Check names, numbers and commitments before relying on the text. Phone calling is not connected.