Workflow guide
From two-hour recording to episode plus clips
Video podcasts must work as full episodes and as a clip pipeline. That demands a different edit: speaker-accurate switching, broadcast-clean dialogue, and chapters that double as clip boundaries. This guide covers the chain from sync to distribution cuts.
From the Mellius product team · Updated September 28, 2026
Conspiracy theorists, you may hate this
AI Director
›
Just type what you need
- ›Remove all silent parts
- ›Add bold kinetic captions
- ›Only have Bill Ackman speaking
- ›Cut to highlights only
- ›Add cinematic transitions
- ›Match cuts to the beat
The short answer
Sync all cameras and recorders to one master audio track, switch video to the active speaker with two-second reaction padding, clean hum and hiss lightly, master dialogue near minus 16 LUFS, then cut vertical clips with fresh hooks from the strongest moments.
Switching feels live
The camera follows conversation naturally, including reactions, not just words.
Dialogue is broadcast-clean
Levels match across hosts and guests with noise unnoticeable at normal volume.
Clips are planned, not scraped
Each episode yields self-contained vertical moments with their own hooks.
Sync once, switch like a director
Sync with waveforms or timecode to a single master audio recorder, then lock that audio and mute camera scratch tracks to avoid phase drift. Multicam switching follows attention: cut to the speaker within a few frames of their first word, hold listeners' reactions for a beat after punchlines, and return to a wide two-shot for transitions and laughter. Resist cutting on every backchannel 'yeah' — it reads as nervous.
Pad switches with roughly one second of anticipation and reaction so cuts do not clip entrances. For remote recordings, align each guest's local file separately since drift accumulates over long episodes; check sync at the start, middle, and end. In Mellius you can ask for speaker-following switching across an episode and correct individual camera choices on the timeline.
- Lock one master audio track; mute camera scratch audio.
- Cut to speakers fast, hold reactions after laughs.
- Use the wide shot for transitions and crosstalk.
- Verify remote-guest sync at three points minimum.
Clean sound and feed the clip pipeline
Level-match every mic before processing so denoise hits all voices equally, high-pass rumble, notch electrical hum, and run gentle voice reduction in light passes. Master full episodes near −16 LUFS stereo so they match platform loudness without listener volume rides. Remove long tangents in a separate pass from um-cleanup — story edits first, polish second.
Mark clip boundaries while editing the episode: self-contained 30–60 second moments with a clear setup and payoff. Each clip needs its own hook in the first two seconds, burned-in captions reframed to 9:16, and enough context to survive without the episode. Export the full episode plus chapters, then batch the vertical clips with distinct titles rather than 'Episode 12 Part 3.'
- Level-match mics before any noise processing.
- Master near −16 LUFS stereo for platforms.
- Mark self-contained 30–60 s moments during the edit.
- Give every clip its own hook and title.
Package episodes for discovery
Chapters, titles, and thumbnails decide whether the edit gets watched. Write chapters around questions and stories rather than timestamps alone, and design thumbnails from a genuine reaction frame with three or fewer words. Show notes should summarize takeaways with links, not transcribe the episode.
Publish consistently sized masters: 4K or 1080p horizontal for YouTube and Spotify video, 9:16 clips for TikTok, Reels, and Shorts. Track which clip topics drive full-episode views and feed those topics back into guest prep. The edit is part of a loop — clips promote episodes, episodes supply clips.
A practical sequence
Put it into practice
- 01
Sync and switch
Lock master audio, align every camera, and cut to speakers with reaction padding.
- 02
Clean and master
Level-match mics, reduce noise gently, and master dialogue to platform loudness.
- 03
Cut and package
Extract self-contained vertical clips with hooks, chapters, and distinct titles.
Frequently asked questions
How do you sync multiple cameras and audio for a podcast?
Clap or use a flash at the top for a sync point, then align waveforms or timecode to one master audio recorder. Mute camera scratch tracks once synced to prevent drift and phasing. For long remote recordings, check sync at the start, middle, and end since files drift apart.
When should the camera switch in a video podcast?
Cut to whoever holds attention: the speaker within frames of starting, the listener for reactions after jokes or bold claims, and the wide shot for transitions, laughter, and crosstalk. Avoid switching on every backchannel murmur. Slightly late beats slightly early — clipped entrances feel broken.
How long should podcast clips be?
Most platforms reward 30–60 seconds with a complete setup and payoff inside the clip. The first two seconds must hook without episode context. Cut several per episode around distinct ideas rather than sequential parts, and title each for its own promise.
Continue reading
Describe the edit. Keep the timeline.
Mellius turns conversational direction into cuts, captions, graphics, motion, color, and sound you can still inspect and change.
Start editing free