Auto cut audio for DaVinci Resolve
Drop the file in a browser, we find the pauses, and you download a timeline Resolve imports — as FCPXML, OpenTimelineIO or EDL, whichever your pipeline wants. No plugin, no Studio licence, and your media is never opened, let alone re-encoded.
In short: set your project’s timeline frame rate before you import anything, upload, check the pauses, download the FCPXML, then File > Import > Timeline. Free tier: 5 min and 200 MB per file, no account.
Three ways into Resolve, and which one to take
Resolve is unusually generous about timeline interchange — it reads more formats than any other editor on this list. We ship three of them off the same analysis. Short answer: take the FCPXML. Long answer, because the other two exist for real reasons:
| FCPXML 1.9 | OpenTimelineIO | EDL (CMX3600) | |
|---|---|---|---|
| What you get | One .fcpxml, media reference included | One .otio — it is JSON | One .edl — it is plain text |
| Carries a media path | Yes, as a file:// reference | Yes, ExternalReference | No. Clip name only |
| Frame rate model | Exact rationals (1001/24000s) | Rate as a number + frame values | Timecode at the rounded rate |
| Readable by a script | XML, awkward | Yes — it is what OTIO is for | Yes, but you parse timecode |
| Take it when | You are importing into Resolve. This is the one | Something else reads the timeline first | The other two failed, or you are handing off to Avid |
The thing nobody tells you: Resolve is not Premiere
Resolve reads FCPXML — Apple’s modern format, root element <fcpxml>. Premiere does not; it wants the old FCP7 XML (<xmeml>), a completely different vocabulary that happens to share a name. Send the wrong one and the import quietly does nothing, which is the classic afternoon-burner. Our export buttons are labelled by editor rather than by format for exactly this reason: you click “DaVinci Resolve”, you get FCPXML, and you never have to know.
The Premiere Pro page goes through the xmeml side of that in detail.
Why we left the format @name out
An FCPXML <format> can carry a name like FFVideoFormat1080p2398. Those identifiers encode a resolution, and we never probe your picture — silencedetect reads the audio stream. So we would have to invent one, and an invented identifier is not a real identifier. We write only frameDuration, which is the attribute that actually drives the conform, and let Resolve resolve the rest from your media. Omitting an optional attribute beats guessing it.
Set the frame rate before you touch anything
This is the one Resolve-specific thing that will cost you real time, and it has nothing to do with us. Resolve locks a project’s timeline frame rate the moment there is anything in the media pool. Import one clip to see if it plays, and the setting greys out.
So the order matters. New project, Project Settings, set the timeline frame rate, then import. And set the same rate in our export dropdown, because we quantise every cut point to whole frames at whatever you pick.
Why we ask you at all
Silence detection runs on the audio stream. We never decode a picture frame, so we genuinely do not know your frame rate and we refuse to guess — a guessed rate produces a timeline whose cuts drift further out of sync the longer it runs. Pick your sequence rate, not the camera’s and not the delivery rate. And the NTSC ones are stored as exact rationals: 23.976 goes into the FCPXML as 1001/24000s and never as a decimal, because rounding it costs about a frame every 40 seconds. Fine at the head of a timeline. Not fine at minute nine.
If it is already locked and you have work in the project, make a new one and import the timeline there. Faster than fighting it.
The import, click by click
1. Project Settings > Master Settings > Timeline frame rate
On an empty project, before anything reaches the media pool. Write the number down; you are about to type it into our dropdown.
2. Upload the file and look at the waveform
Auto-calibration picks a threshold from your file’s own noise floor and highlights every stretch it is going to remove. Drag it and re-analyse if it looks timid — it usually is, on purpose. Correcting it here takes two seconds; correcting it after the import takes a re-import.
3. Pick the frame rate, click “DaVinci Resolve”
You get a
.fcpxmlof a few kilobytes, instantly — no render, no queue. It is available as soon as the analysis is done, so you never have to run the MP4 export at all.4. File > Import > Timeline > Import AAF, EDL, XML…
Pick the
.fcpxml. In the dialog, tick “Automatically import source clips into media pool” and point it at the folder holding your original file. That one checkbox is the difference between a timeline that resolves itself and forty offline clips.5. Relink if it still asks
Our media reference is a bare filename, not a path — see below. If Resolve cannot find the file, right-click the clip in the media pool, Relink Media, point at the original once. Every clip references the same asset, so one relink fixes the whole timeline.
6. Do the part a detector should not do
The cuts are straight cuts: audio and picture land on the same frame. Roll the audio edges where a join sounds abrupt, drop short crossfades on the harsh ones, restore any pause that was carrying meaning. This is the argument for a timeline over a flattened MP4 — you get to disagree with us, edit by edit.
Why every clip arrives offline
Your file reached us through a browser upload and a browser upload strips the disk path. We never knew where your media lives, so the FCPXML says <media-rep kind="original-media" src="file://localhost/your-file.mp4"/> — a bare filename, which always shows offline and always offers a relink. That is deliberate. The alternative is inventing a plausible absolute path, and a wrong path either fails with no explanation or, on a machine that happens to hold a similarly named file, links the wrong media and lets you cut for an hour before you notice. An honest relink beats a silent mislink.
Why we are not a panel inside Resolve
Some of our competitors run natively inside Resolve, as a script or a panel next to your timeline. When that works, it is a better experience than exporting a file and importing it. It is fewer clicks and it knows things about your project that we never will.
It also has to be installed, and it has to keep up with Resolve. Ours does not. A rented suite, a colourist’s machine you have no admin on, an old Resolve nobody is updating mid-delivery, a render node — a text file imports into all of them. And we cover Premiere, Final Cut and Avid off the same analysis, which nothing living inside Resolve can do.
There is a second thing that falls out of being outside. Removing silences by exporting a media file means someone re-encodes your footage. Our own MP4 export does exactly that: H.264 at CRF 20, AAC at 192 kbps. The timeline path cannot — it is string templating over a list of numbers, so no ffmpeg starts and your file is never opened. “No quality loss” is not a claim on this page. It is a property of handing you a text file.
5 min · 200 MB · 1 file/day · no account
60 min · 2 GB · no daily cap
The timeline export is not a paid feature · no watermark · files deleted after 1 hour
The limit is on the upload, never on the output. But 5 minutes is tight — it will not hold an interview — so if this becomes part of how you work, you will be on Pro. See what Pro costs.
Where this stops being the right tool
The detector measures loudness. That is a thirty-year-old idea with thirty-year-old limits, and they are worth knowing before you upload rather than after.
Your noise floor is close to your voice
Then no threshold separates them. On our public benchmark, an Apollo 11 comms tape removed 0.01 s out of five minutes at the auto setting, stayed at zero all the way to −28 dB, then removed 90.3% at −27 dB — one decibel, nothing to almost everything, with nothing usable in between. Tape hiss fills every pause. Denoise first, or use a transcript-based editor.
You want filler words gone
An “um” is sound. We keep it. There is no threshold that finds it, because loudness is the wrong measurement for the question. That needs a transcript, and Resolve has grown transcript tooling of its own.
You need J-cuts and L-cuts
We write straight cuts only — audio and picture on the same frame. Offsetting them so a cut lands under the previous shot's sound is the thing that separates an assembly from an edit, and our serializer lays both tracks in lockstep. You can roll the edges yourself once the timeline is in Resolve. It is on our list to do in the serializer, since it is offset arithmetic over data we already have.
You have one cut to make
Then make it. Topping and tailing a track, or pulling one quote out of an interview, is two drags in Resolve and you already know where they go. We are only worth the round trip when there are too many pauses to click through.
Defaults: threshold auto-calibrated to your file’s mean volume minus 5 dB, clamped to −50…−25 dB, 0.8 s minimum silence, 150 ms of air handed back at each end. The main page has all 155 measurements behind those numbers, on five public-domain files you can re-run yourself, and the Premiere page has a starting-settings table by recording type plus a glossary of the vocabulary.
Cut a file and take the timeline
First one is free, no account. The timeline download appears as soon as the analysis finishes — you never have to run the render, and the FCPXML costs nothing extra.
Questions people actually ask
Which format should I import into Resolve — FCPXML, OTIO or EDL?
▾
FCPXML, unless you have a reason. It carries the media reference, the source duration and the frame rate as exact rationals, so Resolve can find your file and conform the cuts without you telling it anything else. OTIO is for when a script is going to read the timeline before Resolve does, or when you are moving between more than two tools. EDL is the fallback that imports about anywhere and carries almost nothing — clip name and timecode, no media path, no rationals. Use it when the other two have failed.
Why does the Timeline frame rate look greyed out in Project Settings?
▾
Because Resolve locks it as soon as anything is in the media pool. This is the trap that costs people an hour: you create a project, import a clip to check it, then discover you cannot change the timeline frame rate anymore. Set the project's timeline frame rate first, on an empty project, and make it match the rate you chose in our export dropdown. If it is already locked, make a new project — that is genuinely faster than working around it.
Do I need Resolve Studio, or does the free version work?
▾
The free version works. We are not asking Resolve to do anything exotic — importing an XML timeline is a base feature, and nothing about our file needs a paid decoder or a neural engine. Which is worth saying because Resolve's own transcript-based tooling and the plugins that run inside it are not all available on the free build.
Does the XML re-encode my footage?
▾
It cannot. The timeline endpoint reads the cut list already sitting in our database and templates a string — no ffmpeg process starts, no worker picks it up, no queue is touched, your media file is never opened. Resolve then plays your original off your own disk. This is the difference between our two exports: the MP4 export really does re-encode (H.264, CRF 20, AAC 192 kbps), and the timeline export mathematically cannot.
Every clip is offline after import. Is that normal?
▾
Yes, and it is deliberate. A browser upload does not carry a disk path, so we never knew where your media lives. Our FCPXML points at file://localhost/your-file.mp4 — a bare filename — which makes Resolve show the clip offline and ask. Relink once against the original and the whole timeline resolves, because every clip references the same asset. Ticking “Automatically import source clips into media pool” in the import dialog and pointing it at the folder handles this in one step.
Can it drive the detection from one mic on a multitrack recording?
▾
No. We analyse the mixed-down audio and write a single audio track, one clip per kept segment. On a two-mic interview that usually behaves the way you want — a gap only counts as a gap when neither person is talking — but you cannot say “cut on the host's mic only”, and if your two tracks have very different noise floors the mix will follow the noisier one. Track selection is a real gap and we have not built it.
Does it remove “um” and “uh”?
▾
No. We measure loudness, and an “um” is sound. Removing filler words needs a transcript, which we do not produce — Resolve has grown transcript-based editing of its own, and Descript built a company on it. If verbal tics rather than dead air are your problem, one of those is your tool. What we give you instead is language independence: silence detection works identically on Japanese, Wolof and English, because it never listens for words.
Can I use it in a Fairlight-only, audio-post workflow?
▾
Yes, and the FCPXML is the wrong choice for that — take the OTIO or drop an audio file in rather than a video. If you upload audio we write an audio-only timeline (no video track at all), which is what you want landing in Fairlight. Bear in mind the cut list comes from a mixdown analysis: it is a rough assembly to work from, not a conform.
What does it cost?
▾
Nothing for the first file: 5 minutes, 200 MB, one file per day, no account. Pro is 60 minutes and 2 GB with no daily cap. The limit is on the upload, not the output — no watermark, no reduced-quality export, and the timeline download is not a paid feature at any tier. Straight talk: 5 minutes will not fit a real edit, so if this becomes part of your workflow you will be paying.
Can I get the cuts without waiting for a render?
▾
Yes, and this is the fastest path. The timeline export is available the moment the analysis finishes — it does not need the export step at all, because it reads the same silence data the render would have used. Upload, check the waveform, download the FCPXML, close the tab. On our benchmark, detection on five minutes of audio took between 0.3 and 0.5 seconds.
No plugin, no Studio licence. Run a file through it.
Five minutes, no account, no card, and your rushes never move.
Cut my file and get the timeline