1. Give the interview a source
Start with the person who will speak and the kind of conversation they are here to have: founder or builder, operator or practitioner, creator or commentator, researcher or expert, or a custom profile.
Add a link to their work and any material the interview should examine. This might be a post, product page, paper, project note, launch recap, draft argument, or list of claims that need a clearer explanation.
The source pack is not a script. It is evidence and context the question writer can use to avoid generic prompts.
2. REC researches and prepares the question path
REC examines the supplied material and relevant public sources, then prepares a path of typically 12–15 questions. The mix depends on the conversation profile and the amount of usable context.
Good prompts ask for something inspectable: a specific example, decision, number, artifact, disagreement, process, framework, or piece of evidence. If an important detail is missing, the question can ask the speaker to provide it on camera.
The reviewed path is prepared before recording. When compatible live transcription is available, REC can quietly prepare one grounded follow-up from the answer in progress; it appears only as an optional cue at a natural pause, so the user still directs the interview rather than handing it to a live host.
3. Check the setup and record real answers
Before recording, the browser checks the available camera, microphone, recording format, and any chosen visual treatment. The camera keeps rolling while the speaker moves through the prepared conversation at their own pace.
Answer anchors can suggest what kind of detail to reach for, but the words come from the speaker. The most useful answers tend to contain one clear claim, one concrete example, and enough context to understand why the point matters.
The session records the speaker rather than generating a stand-in. That makes the full recording the source of record for everything that follows.
4. Browser support determines automatic highlights
Compatible browsers can provide live speech-recognition segments during the recording. REC uses those timestamped segments to look for self-contained, specific, and potentially useful moments.
The highlight selector aims for several moments that are usually short enough to stand alone. Each suggestion includes start and end timestamps, a label, and a rationale so the user can judge whether the excerpt deserves to become a clip.
If live transcription is not supported, REC skips automatic highlight recommendations. Recording still works. This is an explicit current product boundary, not a hidden quality setting.
5. Choose, render, and share manually
The user chooses a recommended moment and a static clip style. REC cuts the source recording to that recommendation, adds the selected title and speaker overlay, and renders the result in the browser.
The rendered clip can be downloaded. The suggested caption can be edited or copied, and a compatible browser can open its native share sheet.
REC does not currently publish, schedule, or choose a social channel for the user. The final editorial and distribution decision remains manual, which keeps approval with the person responsible for the claim.