To turn a podcast into shorts you need the video recording, a way to find the 30 to 90 second moments that stand on their own, and a pass that reframes each one to 9:16, captions it word by word, trims the filler and puts a hook title on screen. Done by hand that is 4 to 6 hours per episode. Done with a clipper built for conversation it is about 10 minutes of rendering and 15 minutes of picking. This guide walks through all seven steps, what each one is for, and where the time goes.
Step 1: Get the video, not just the audio
Shorts are a video format. Export the episode as MP4 or MOV from your recorder (Riverside, Zoom, StreamYard, a camera) or use the YouTube link if the episode is already up. Two things about the source decide what you can do later:
- A wide shot with every speaker visible is what makes multi-speaker split-screen possible. If your recording only has solo close-ups, you get active-speaker framing instead, which is fine but loses the reaction shots.
- 1080p is enough. A 2 hour 4K file is mostly wasted bandwidth for a 1080x1920 output.
Step 2: Find moments that are complete, not just loud
This is where most podcast clips fail. The tempting shortcut is to look for the loudest, most animated 45 seconds. What actually works is a segment with three parts:
- Hook: the first sentence makes a claim or asks a question.
- Context: enough setup that someone who never heard the episode follows it.
- Delivery: it ends on a resolved thought, not a trailing “so anyway”.
By hand, the honest way to do this is to read the transcript and mark paragraph boundaries, which takes 30 to 45 minutes for a 90 minute episode. Skapo does the same thing semantically: it reads the transcript, scores each segment on hook strength, context completeness and clean ending, and hands you the candidates ranked. You still pick; you just pick from a shortlist instead of a timeline.
Step 3: Cut on speech boundaries
A clip that starts half a word in, or ends with a pop where a syllable was cut, reads as cheap before anyone has processed what was said. Put the in and out points on natural pauses. In a manual editor that means zooming into the waveform for every cut. Skapo snaps every boundary to the nearest pause automatically, which is why its clips do not open mid-word.
Step 4: Reframe to 9:16 without losing a speaker
A 16:9 podcast frame cropped to vertical keeps one third of the picture. For a solo speaker, keep them centred and let the crop follow them. For an interview, stacking both people in a split-screen is what stops the guest vanishing every time the host talks. Skapo detects 2 to 4 speakers from the wide shots and builds the split layout before rendering, so the frames never jump when someone leans in. The mechanics are on the split-screen page.
Step 5: Captions, timed per word and out of the way
Most shorts are watched muted for the first second, so captions are the hook as much as the audio is. Two rules: time them per word, not per sentence, and keep them out of the bottom 20 percent of the frame where the TikTok, Reels and Shorts buttons sit. If your show is not in English, check that the tool renders the script at all; until August 2026 most clippers, Skapo included, burned Japanese and Thai captions in as empty boxes. Skapo now ships fonts for those scripts and tunes 22 languages end to end, listed on the languages page.
Step 6: Trim filler and dead air, in the right language
Every “um”, “so” and half-second silence you remove makes a 60 second clip feel like 45. The catch is that filler is language-specific: “er” is hesitation in English and a pronoun in German, so an English filler list applied to a German episode cuts real words. Use a per-language list, or a tool that has one.
Step 7: Title, cover, export, schedule
- Put a 3 to 6 word hook title on screen for the opening seconds. It is the strongest single lever for stopping the scroll.
- Pick a cover frame from the hook with the title on it; libraries of 30 clips are unreadable without covers.
- Export at 1080x1920, H.264, under 60 MB for TikTok’s upload path.
- Post one clip a day from the episode rather than ten at once. The platforms reward cadence, and you learn which topics land.
Where the time goes: manual, generic clipper, Skapo
| Step | By hand | Volume-based AI clipper | Skapo |
|---|---|---|---|
| Find moments | 30 to 45 min transcript read | Automatic, picks loud spikes | Automatic, ranked by story completeness |
| Cut and reframe, per clip | 15 to 25 min | Automatic, often mid-word | Automatic, snapped to pauses |
| Two speakers on screen | Manual layout per clip | Usually single-speaker crop | Split-screen, 2 to 4 speakers |
| Captions, per clip | 10 min plus fixes | Automatic, English-first | Automatic, 22 tuned languages |
| Wait for render | Real time export | 20 to 40 min queue at peak, per OpusClip’s own help center | About 10 min for the whole set |
| Full episode, 8 clips | 4 to 6 hours | 45 to 90 min including fixes | About 25 min including picking |
Manual timings are what agencies report in our onboarding calls; Skapo’s are the production median. OpusClip’s queue figure is from their published help documentation, and it varies with load. If you are choosing between tools, the OpusClip comparison says plainly when each one is the better fit.
Try it on your own episode
Paste a YouTube link or upload the file on the try page, no card needed. The free tier processes two hours of source video a month with watermarked exports, which is enough to see whether the moments it picks are the ones you would have picked. Podcasters specifically get the workflow on the podcast page.
The timings in this guide are medians from real jobs. The full distribution, including the slow tail and the method behind it, is on the production benchmarks page.
Charging clients for the clips you make? The rate benchmarks are in the 2026 agency pricing guide.