What you need
- Claude Code in a terminal, pointed at a folder with your footage.
- ffmpeg and ffprobe on your PATH.
- A local Whisper model for word-level transcripts.
- Python with numpy, soundfile and OpenCV for the checks.
Find every pause
ffmpeg silencedetect lists each stretch below a noise floor for longer than a set time. A 0.4 s minimum catches dead air without touching natural phrasing.
ffmpeg -i work/episode_16k.wav -af silencedetect=noise=-35dB:d=0.4 -f null - 2>&1 | grep silence_Set the floor from your own recording. Claude reads the room tone off the first seconds of the file and puts the threshold about halfway between room and speech. A fixed -35 dB is a starting point, not a rule.
Trim to a breath, not to zero
Cut each pause down to 0.15 to 0.25 s. Zero-gap cuts read as robotic, and a real beat before a punchline stays.
List the pauses
Run silencedetect on the 16 kHz track and keep every gap over 0.4 s.
Keep a breath
Shorten each gap to about 0.2 s, centred on the quietest point. Leave the pause before a punchline alone.
Render the keep list
trim and atrim each kept range, 15 ms fades at each join, concat.
Hide the jump
On one camera, alternate a gentle punch-in at each join, or a 5-frame dissolve on horizontal edits.
Check sync
Audio edits stay at their exact time; only the picture snaps to whole frames. Snapping the audio too shaves up to 20 ms off a word.
ffmpeg -i episode.mp4 -filter_complex "\
[0:v]trim=3091.42:3094.87,setpts=PTS-STARTPTS[v0];\
[0:a]atrim=3091.42:3094.87,asetpts=PTS-STARTPTS,afade=t=out:st=3.435:d=0.015[a0];\
[0:v]trim=3096.12:3099.40,setpts=PTS-STARTPTS[v1];\
[0:a]atrim=3096.12:3099.40,asetpts=PTS-STARTPTS,afade=t=in:d=0.015[a1];\
[v0][a0][v1][a1]concat=n=2:v=1:a=1[v][a]" -map "[v]" -map "[a]" out/clip.mp4When a jump cut works and when it does not
Vertical shorts can cut hard. Horizontal edits for YouTube and LinkedIn read better with angles held 7 to 9 seconds and soft dissolves over the tightening jumps.
- Vertical (TikTok, Reels, Shorts): faster, punchier cuts are fine.
- Horizontal (YouTube, LinkedIn, X): hold each angle 7 to 9 s and hide tightening jumps with a ~5 frame dissolve on the same shot.
- Never a visible jump on the same framing at the payoff. The last 5 seconds of speech stay on the speaker.
Every Claude video editing guide
Frequently asked questions
How do I remove silence from a video with ffmpeg?
Run silencedetect to list the gaps, then trim and concat the ranges between them. Claude Code writes the filter graph for every range so you do not hand-type timestamps.
Is there a free jump cut remover?
ffmpeg and Whisper are free and open source, and Claude Code drives them on your own machine. You pay for Claude, not for a per-minute export.
How long should pauses be after editing?
About 0.15 to 0.25 seconds. Anything over 0.4 seconds reads as dead air on social video; zero reads as robotic.
Want finished ads, not editing sessions?
Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. Strategy, hooks, scripts, shooting and editing, end to end. You never write a prompt. You pick the winners. Sprites ships the campaigns. In private beta.