AI Video Editing Flows

Silence remover tools compared on one talking-head video

One talking-head video goes through auto-editor, video-use, TimeBolt, Gling, Recut and other silence removers from our research list. For each tool we report the time to an accepted export, cuts in the wrong place, pauses left behind and cost. Results are published only once the run is complete.

This page is built but kept out of search until it carries the evidence it needs: No measured results yet; needs a recorded talking head we may use.

Status: in testing

In testing. Nothing on this page has been measured yet, and it shows no result until a test has been run and reviewed.

Nothing has been measured yet. This page describes a test we plan to run, and it will stay without results until the test has been run and reviewed.

The question this page answers

On the same talking-head video, which silence removal tools get to an accepted export fastest, and which ones cut in the wrong place? We have not run this comparison, and this page is the plan for it.

The protocol

The job is to cut the dead air from one talking-head recording and export a file you would be willing to publish.

  1. Record or obtain one talking-head video with speech that we have the right to publish. Mark by hand where every pause longer than 0.5 seconds begins and ends. That marked list is the reference.
  2. Run each tool on the same file. We plan to compare auto-editor, TimeBolt, Recut, Gling, Descript and the silence remover on this site. The browser tool works on audio only, so it is judged on the audio cut list.
  3. Use each tool's default settings first, then the settings its own documentation suggests. Write down every setting.
  4. A person reviews each result and fixes it until it is acceptable. That person's time is counted.
  5. Compare each tool's cuts with the reference list.

To rerun it, take the same file, the reference list and the settings we publish.

What we will measure

  • Time from opening the file to an accepted export, including the reviewer's fixes.
  • Cuts in the wrong place: cuts that clip the start or end of a word, or land mid-sentence.
  • Leftover pauses: pauses longer than the threshold that stayed in.
  • Cuts that remove wanted pauses, such as a beat before a punchline.
  • For paid tools, the plan needed and its price on the date we checked it.

Each measure is reported separately for each tool. We do not make one overall score.

What we need to run it

We need a recording of a person speaking, with the right for us to publish it and the test results next to it. We have not recorded any footage yet. No clean footage set exists for this, which is why we plan to record our own.

We also need a controlled machine, so the timing is not affected by other work. Paid tools need a paid plan or trial for the period of the test.

Method and limits

No results appear here until the comparison has been run on the shared video.

A single talking-head video cannot cover every speaking style, so the results will describe that footage only.

Settings differ by tool, so we record the settings used for each run. Our method is on How we curate and test AI video workflows.

The source is our own footage under an open licence, so anyone can rerun the test. Every tool gets the same acceptance rule: an export is accepted when a person finds no clipped words and no distracting pauses. Human finishing time is counted.

Page checked on Nov 4, 2026.