Podcasts
Turn one podcast episode into the clips worth posting.
An hour of conversation usually holds three or four moments that stand on their own. Finding them means listening to the whole hour again, with a timeline in front of you. ClipSpark does the listening and hands you a ranked shortlist with the reasoning attached.
Why episodes are hard to clip
The good part is in the middle
A strong podcast moment is usually an answer, which means the sentence that sets it up is thirty seconds earlier. A clip that starts on the punchline has no context, and one that starts too early loses the viewer before it arrives.
Nothing marks it
A conversation has no scene changes and no cuts. The waveform is a flat wall of speech, so the visual shortcuts that work for edited video tell you nothing about where the interesting part is.
It is shot wide
Most podcasts are recorded 16:9 with two people in frame. Cropped to 9:16 down the middle, you get half of each person and neither of them properly.
It is watched on mute
A clip that autoplays silently in a feed has to work without sound for the first second or two, which a talking-head crop of a conversation does not do on its own.
What ClipSpark does about each of those
The whole episode is read
Not a sample of it. A moment ninety minutes in is as findable as one in the first five, and every candidate arrives with the hook line and the reason it scored — including the ones that were considered and dropped.
Cuts are placed, not trimmed
You choose a length band of 15–30, 30–45 or 45–60 seconds and the cut is positioned to fit it — so the setup is inside the clip rather than sacrificed to a fixed duration.
The crop follows the speaker
Optional, and the setting that matters most on a two-person show. The 9:16 frame moves onto whoever is talking and switches when the other person takes over, instead of holding one framing for the whole clip.
Subtitles are burned in
Generated from the episode's own audio and rendered into the file, in one of three looks chosen to stay legible on a phone. They survive being reuploaded anywhere, because they are part of the picture.
Settings for a conversation
Pick the audience
Podcast and Interview are both in the list, and they rank differently. An interview episode's best moment is often the guest's unplanned answer; a two-host show's is more often the exchange.
Pick what counts as good
Funny, emotional, educational, controversial, strongest hook, most shareable, or simply exciting. This changes which moments win, not just the caption written for them.
Cut the dead air
Optional silence removal tightens the clip without touching the short pauses between sentences — those are most of what makes speech listenable, so they are left alone deliberately.
Ask for several at once
One upload can produce a batch of clips rather than one, and each extra candidate costs a few more credits instead of a second full charge for the episode.
What to know before you upload
It needs the video recording. ClipSpark reads the picture as well as the audio, and an upload with no video track is rejected at the door — an MP3 with cover art included. If your show is audio-only, there is nothing here for you yet. If you record video to camera and publish audio, the camera file is the one to upload.
Speaker following needs faces it can see. It works out who is talking by watching mouth movement on camera, not by analysing the audio, so it wants faces that are visible and roughly front-on. When it cannot tell — nobody facing the camera, or two people talking over each other — it holds one framing rather than guessing. It is free either way.
MP4, MOV, WebM or MKV, up to 2 GB. H.264 is the safe export. Your plan sets the maximum episode length; credits are charged by length on a published table, and a job that fails or is cancelled returns them automatically.
Every feature in full → What an episode costs in credits → Clipping a stream instead →
Start with one episode
Upload the last one you published.
The free plan is enough to see whether the cuts are any good.