Edit to the beat: how to sync video to music, with the frame math

What is a beat marker?

A beat marker is a timeline marker placed at the time of a beat in a song, so that cuts, trims and clips can snap to it. Three words matter before you start, and most guides skip them.

• Beat: the steady pulse a listener taps along to, per Wikipedia's entry on the beat.
• Downbeat: the first beat of a bar, number 1, usually the strongest accent. In 4/4 time it comes every fourth beat.
• Phrase: a run of bars that sounds like one musical sentence. We found no standard length, so count it by ear: listen for where the pattern repeats or the music changes, and note the bar where that happens.

Editing to the beat means choosing which of these your cuts land on. That choice shapes the feel of the piece more than any detection tool does.

Should you cut on every beat, the downbeat or the phrase?

Cut on fewer beats than you think. Landing a cut on every beat is exhausting after a few seconds, and at higher tempos the shots go by too fast to read. This is craft advice, not a measurement. For why rhythm is one of the criteria editors weigh, see the page on editing rules.

One rule of thumb from Rushilo's skill file: under about 0.4 seconds a picture cannot be read. That is our guideline, not a study. At 150 BPM a beat lasts exactly 0.4 seconds, so cutting every second beat (0.8 seconds) keeps each shot watchable. Also remember that the visible hit, such as a door slam or a foot landing, is what should meet the beat. The cut itself often looks better a frame or two before it.

How many frames is a beat?

Frames per beat equal fps times 60 divided by BPM. That is the same as fps multiplied by the seconds per beat, where seconds per beat is 60 divided by BPM. Frames per bar in 4/4 are four times that number.

Worked line: 120 BPM at 30 fps. 60 divided by 120 is 0.5 seconds per beat. 30 times 0.5 is 15 frames per beat, 60 frames per bar and 240 frames (8 seconds) for four bars.

The half-frame case

A cut can only land on a whole frame. At 128 BPM and 24 fps a beat is 11.25 frames long, so a beat-locked cut can sit up to half a frame from the true beat. Half a frame is 0.5 divided by the frame rate: about 21 milliseconds at 24 fps. Almost nobody can see that, and Resolve snaps to the nearest frame for the same reason. It only matters if the error grows, which brings us to drift.

The 29.97 versus 30 drift

If a beat grid is calculated at 30 fps but the project plays at 29.97, the two disagree by 0.03 frames per second. Over a 3 minute song that is 180 times 30 times 0.001, or 5.4 frames, enough to see a cut arrive late. The same gap between 23.976 and 24 is about 4.3 frames over 3 minutes. Whether this happens depends on how your editor stores time, so treat it as a check: land a cut near the end of the song, play it, and see if it still hits.

How do CapCut, Premiere and Resolve do it?

Everything in this section was checked on October 22, 2026. Features and menu names change, so confirm in your own version.

CapCut

CapCut's help page says Auto Cut is available on mobile and desktop, not CapCut Web. On desktop it lists Beat Sync, described as for music-driven edits, under AI Tools. On mobile the music-beat option sits inside Auto Cut and may be labelled Beat Cut or Rhythm Edit. A search overview also describes beat markers on the audio track, but we could not confirm those labels on a CapCut page, so look in the app. Pricing and paywalls are covered in the CapCut paywall page.

Premiere Pro

We found no built-in beat detection in desktop Premiere Pro as of October 22, 2026; an Adobe community request for it was open. The manual method is to select the song, play it, and press M on each beat to drop a marker, then snap clips to those markers. Plugins such as BeatEdit or Beat Detector automate it, and users report prices around $100. Premiere on iPhone is different: per Adobe Help you select the audio track, tap Beat, switch it on and choose Fewer, More or All.

DaVinci Resolve

Blackmagic announced AI Detect Music Beats with Resolve 20 on April 4, 2025. It analyses beat-driven music, places markers, and lets clips snap to them. A third-party guide, which cites the Resolve manual, says it needs the paid Studio edition and suits 4/4 or 3/4 music. We did not read the manual, so check Blackmagic's feature list for your edition. Without it, press M on the music clip.

What does beat detection get wrong?

Detection finds a pulse. It does not understand the song. Expect these limits from any tool, and check the markers by ear before you cut to them.

• Tempo range. Detectors search a window of tempos. Rushilo searches 60 to 200 BPM, another tool we read states 50 to 220. Music outside the window can come back at double or half the real tempo.
• Downbeats are guesses. Telling beat 1 from beats 2 to 4 is harder than finding the pulse. Songs in 3/4 or ones that start with a pickup can put downbeats on the wrong beat.
• Tempo changes. A fixed grid drifts when music speeds up, slows down or plays free-time passages.
• No pulse, no beats. Speech, ambience and rubato music may return nothing.

How does an agent in Rushilo cut to the beat?

We make Rushilo, so weigh this section accordingly. Rushilo is a desktop editor for Mac and Windows that your own agent (Claude Code, Codex or another MCP client) drives; it ships no AI of its own. Version 1.3.0 added beat finding and a cut-to-the-beat skill, per the release notes. Everything below comes from reading Rushilo's code. We did not run the beat finder on real music for this page, so we make no accuracy claim.

What the beat tool returns

The rushilo_beats tool runs on your computer with no model and no cost. Given a music clip or track, it returns one tempo (an average over the clip, or none), the beat times, and the downbeats, which it takes as every fourth beat, assuming 4/4 and starting at whichever position carries the most onset strength. Its own answer calls the downbeats a best guess. It searches 60 to 200 BPM and needs at least 3 seconds of audio; with less, or with no steady pulse, it says so and you place cuts by hand.

Unless told otherwise, it writes ruler markers named Beat n and Downbeat n in one undo step. Re-running replaces the earlier beat markers, markers you named yourself stay, and the Clear beats button removes only the ones that match those names.

How cuts snap

Beat markers are ordinary markers. In the editor, a drag, trim or move snaps to a marker within 8 pixels at the current zoom, so the reach in time shrinks as you zoom in. That is why hand edits stay on the beat after the agent has laid out the cut.

The agent route

1. Import the music, then trim it first, because beats are found in the clip as it plays.
2. Run rushilo_beats and read the tempo.
3. Choose a cut rhythm from the tempo and say why: downbeats, every second beat or every beat.
4. Lay the whole layout as one batch of clip placements, each starting at a beat time, so one undo reverses it.
5. End on a downbeat, or fade the music out.
6. Check a few cuts with rushilo_frames and correct drift.

A whoosh from rushilo_sfx can sit on a cut time (see adding sound effects), and a music bed under speech should be ducked, which this guide covers.

For the agent setup see the skills overview, or Buy Rushilo.

Worked example: a 90 second promo at 128 BPM

These numbers are constructed for this page, not taken from a real project. The song is 128 BPM, the project is 30 fps, and the cut rhythm is every second beat.

A beat is 60 divided by 128, or 0.46875 seconds, which is 14.0625 frames. Every second beat is 0.9375 seconds, or 28.125 frames. A downbeat comes every four beats: 1.875 seconds, 56.25 frames. Over 90 seconds that is about 96 cuts at this rhythm.

Only some of those land on a whole frame, so the editor rounds each to the nearest one, an error of up to half a frame (about 17 milliseconds at 30 fps).

Common mistakes

• Cutting every beat at 150 BPM, so shots last under half a second.
• Cutting the action instead of the hit. Line up the impact with the beat, not the edit point.
• Trusting downbeats on a waltz or a song with a pickup. Listen first.
• Finding beats before trimming the song, so the markers are on the wrong clip positions.
• Ignoring frame rate, then wondering why the end of the song is late (see the 29.97 note).

For how a hard cut differs from a jump cut, see the jump cut page. We did not test beat accuracy in Rushilo, CapCut's labels, or export-time frame rounding of cuts, and this page will be updated as vendors change.

Patel, O. (2026, October 22). Edit to the beat: how to sync video to music, with the frame math. Rushilo. https://rushilo.com/blog/sync-video-to-beat

About the author

Om Patel

Founder, Rushilo

Om Patel is the founder of Rushilo, a desktop video editor driven by the user's own AI, and of BigIdeasDB, a research platform that reads real complaints and reviews to help founders validate products. He writes Rushilo's research articles with the same method: primary sources, dated numbers and the words real users write.

• X (Twitter)
• BigIdeasDB