Cropping a widescreen video to vertical keeps the middle of the shot: the table, the mic, someone's shoulder, while the person talking slips off the edge. Fitting the whole frame makes everyone tiny. What you need is a frame that moves with the conversation, and doing that by hand in an editor takes hours.
Upload a video up to 10 minutes: an interview, a podcast, a talk, a conversation on camera. ElevenLabs Scribe transcribes the speech and labels the speakers, then AI finds the faces in the shot and matches them to the voices. In Auto mode, whoever is speaking fills the frame, two people who talk over each other share a split screen, and wide shots and slides are shown whole over a blurred background. If the speaker's face is hidden behind a mic or a voice can't be matched to a person with confidence, the video is shown whole too, so nobody ends up cut off. You can also pick a mode yourself: always follow the speaker, two on screen, or the whole frame.
Captions are one switch away: five styles, with words appearing in time with the speech. You get a 1080×1920 MP4 with the original sound, ready for Reels, YouTube Shorts and TikTok.
The tool works on the web and in our mobile app. If processing fails, your credits come back. To pull several clips out of a long video, use the AI Clip Maker instead.