How to turn a podcast or interview video into vertical shorts (with face tracking)

A one-hour video podcast usually has a handful of moments that would work as 30-to-60-second vertical clips. The tedious part is the reframing: a 9:16 frame only covers about a third of a 16:9 picture, so you have to keep moving it to follow whoever is talking, and then add captions because most people watch shorts with the sound off.

This guide uses TemaShort, a Windows app from TemaLab that does the reframing and captions for you. We make TemaShort, so read with that in mind. The steps work on the free version (exported shorts have a watermark until you add a license key).

Use it on videos you have the right to clip: your own show, your streams, courses, webinars and talks. If guests appear, make sure they're happy for clips to be posted, and don't clip other people's videos without permission.

1. Mark the clips

Install TemaShort and open your episode (MP4, MOV, MKV or WebM). In the Clips step, play the video and press “From here (I)” where a clip should start and “To here (O)” where it should end. Each clip becomes one short, and you can add as many as you like. You can also add parts of another video as extra scenes and join them into one short.

2. Let the frame follow the face

In the Framing step, TemaShort analyzes only the clips you marked, finds the faces, and moves the vertical frame with the person shown longest and largest. The blue box on the left is the 9:16 crop; the preview on the right shows the result.

• Wrong person, or the face was lost? Press “Follow another from here” or “Hold frame from here” at that point in the timeline, then drag the frame if needed
• Zoom can be Normal, 1.25× or 1.5×. Cropping 9:16 from a Full HD video enlarges a 608×1080 area to 1080×1920, so it's a little softer; 4K sources stay sharp
• Profiles, dark shots and very small faces can be missed — fix those spots by hand

3. Two people: stack them or switch to the speaker

Interviews and co-hosted shows have two people in the shot. You have two options.

Stack them. Choose “2 stacked” to put one person on top and the other below. Both stay visible the whole time, which suits fast back-and-forth conversation.

Switch to whoever is speaking. Keep “Follow 1” and press “Switch to the speaker automatically”. TemaShort tells the speakers apart by voice, matches each voice to a face, and adds a switch mark where the speaker changes. You can change the matching and edit any mark; similar voices or crosstalk can switch at the wrong moment.

Under Trim, you can also cut silences (trim pauses in speech) and cut parts where nobody you're following is visible.

4. Add captions and a title

In the Subtitles step, press “Add subtitles (transcribe)”. TemaShort transcribes the speech in each short on your PC (the first time, it downloads the transcription model, about 574 MB). Set the spoken language to English, Japanese or Chinese, or leave it on automatic.

• Edit the text and timing in the list; split, join or shift lines
• Pick one of four templates and three sizes, and drag the captions up or down. “Show areas that may be covered” marks where platform buttons tend to sit
• Add a title bar of up to two lines at the top (“Ep. 12: Plan your week in 3 lines”)
• Export the captions as an .srt file if you'd rather upload them separately

On our test recordings made with synthetic voices, the English captions had an average word error rate of 2.3%. Real recordings have more errors, so read through the captions before you export — especially names and numbers.

5. Export

In the Export step, choose the shorts and a folder, and TemaShort writes vertical 1080×1920 MP4 files (H.264/AAC, original frame rate up to 60), named “<original>_short_01.mp4” and so on. Existing files are never overwritten. Upload them with each platform's own app or website.

The free version adds a “TemaShort” watermark in the middle of exported shorts; the on-screen preview and .srt files don't have it. With a license key, there's no watermark and no limit on how many shorts you make or how long your videos are.

6. Tips for better vertical clips

• Start on a sentence. Begin each clip at the start of a thought, not mid-sentence, and end after the point lands
• Keep it short. One clear point per clip. TemaShort warns you if a short goes over three minutes, because some platforms don't accept longer shorts
• Record with vertical in mind. Leave some space above heads, and keep two-person shots wide enough that each face can fill a 9:16 frame
• Let Claude do the repetitive part. With the full version, you can tell Claude Code or the Claude desktop app the timestamps and get captioned shorts. See How to have Claude turn a long video into vertical shorts

Published: October 27, 2026 · TemaLab