✂️Remove silences

❓ What is it?

Almost every edit starts with raw footage that has useless, silent portions. Remove silences cuts all these silences out with only a few clicks.

🤔 How does it work?

Silence cutting in FireCut is powerful and flexible.

  • You specify which audio tracks to use as guidance tracks

  • When all guidance tracks are silent, that portion of your sequence is cut out i.e., all tracks are cut. This is because you might have your "good" audio on many different tracks / clips, and they might not even by linked to your video. This does not matter for FireCut, as long as you select the right guidance audio tracks

For your first silence cut, you should apply it to a small part of your sequence (5 mins) to see what it actually does.

⚒️ Basic vs. Advanced mode

  • Basic: One-click operation with 1 main setting: cutting tightness. If you want some short gaps before and after silences in the result, then choose a loose cut. If you prefer a fast-paced video with very sharp cuts, then pick a tight cut. When you click "Go", FireCut will detect silences and cut them out, all in one go.

  • Advanced: Three-step operation to allow greater customisability:

    1. Detect silences based on various settings

    2. Review silences in a list, and customise which ones to keep, how much padding to add

    3. Cut out these silences

If you're cutting silences from a large sequence (e.g. 1 hour of footage), you should use Advanced mode so you can preview and fine-tune the output before making the cuts

⚙️ Settings

Both modes have:

  1. Silence guidance audio tracks: The audio tracks you want FireCut to listen to for silences. Whenever all of these tracks are silent, that portion of the sequence will be marked for cutting.

  2. Scope: You can decide whether to run the operation on your Full sequence or only on a certain portion that you specify with In / Out points (you can place these in your sequence using the I and O hotkeys)

  3. Speech detection: Experimental feature which detects portions of your sequence which are not silent, but might not have any words being spoken in them. This feature is for removing background noise which is above the silence detection threshold.

  4. Cutting algorithm: How FireCut performs the cuts (this does not affect where the cuts actually happen). Over time, we release newer and faster algorithms. Currently we have:

    • Original: Most reliable but slow, best for short sequences (<10 mins scope, <50 silences)

    • Rapid: Faster and fairly reliable, best for longer sequences (10-30 mins, 50-100 silences)

    • Turbo: Fastest but quite experimental as of late July, best for very long sequences (30 mins to multiple hours, 100+ silences)

If you're using Basic mode, there is just 1 main setting that determines how silences are identified: the silence cutting tightness. You should start with a loose cut, and repeat if you need for tighter cuts. After a few tries, you'll find a value of tightness that works well for you!

For Advanced mode, there are a few extra options on detecting silences:

  • Silence threshold: This is the audio level in dBFS (decibels relative to full scale) below which FireCut will deem your sequence to be silent. If you're not familiar with dBFS, it is the same scale you see in the Premiere Pro audio level indicator. It has a maximum value of 0 dBFS (which means audio levels are usually negative). The right threshold depends on your audio source but a good starting point is -30 dBFS

  • Auto-detection of silence threshold: FireCut can recommend you a good threshold to use. It does so by analysing the audio of your sequence (the portion in scope) and plotting the frequency of different loudness levels. These are then shown in a small chart under the threshold slider. In this chart you will typically see two peaks:

    • The peak on the right (at a higher dBFS loudness level) is your useful audio e.g., your speaker, and will likely be around -20 dBFS

    • The peak on the left (at a lower dBFS loudness level around -50 dBFS) represents the silent portions in your scope

    Note: This chart shows how common the different loudness levels are, so for example if you are speaking half the time at -20 dBFS, and are silent for the remaining half of the time with a background noise of -50 dBFS, then you will see two peaks in this chart of roughly the same size at -20 and -50 dBFS. You should pick a threshold between these two values (such as -35 dBFS) so that FireCut picks out only the portions below that threshold to be silences.

  • Minimum silence duration: This is the minimum duration for which your audio needs to be below the threshold in order to be detected as a silence portion that should be cut. While FireCut's flexibility allows very small values, this is not recommended because it will create lots of small cuts even at breaks between words, which is usually undesirable! A good starting point is 750ms

For Advanced mode, once silence detection is done, you can review and modify the list of detected silences:

  • Cycle through silences: You can click on any of the silences (or the next / previous silence buttons) to see in your timeline. Clicking the "X" next to any silence will remove it from consideration, and it will not be cut.

  • Padding: If you find that the silence detection was too "aggressive" (e.g. it's marked the start/ends of words as silences), this can be fixed by increasing padding. Padding is basically just extra space you allow for your clips. Increasing it has the effect of pretending the silence was shorter than what was detected, and so it leaves some extra breathing room between cuts. Some detected silences might not be cut at all when you increase padding (because the padding is more than the duration of the silence). A good starting point for padding is 250ms on start AND 250ms on end

  • J-cuts offset: Increase this above 0 frames to create J-cuts (i.e. cutting the audio earlier than the video). As the diagram in this review screen shows, the offset is actually applied to the video clip, not the audio clip, so your silences will still be cut in the same places. This is done to ensure you don't accidentally cut the starts/ends of sentences. More details on J-cuts here.

💡 Tips for best outcomes

  • Get familiar with Advanced mode early. It offers much greater flexibility!

  • Don't use settings that will create hundreds of cuts, which are often undesirable and will take too much detection / processing time. For example:

    • Don't set the detection threshold (dBFS) too high. Going over higher than -30 dBFS is not recommended

    • Keep the minimum silence duration above 750ms

    • Always add some padding before cutting, this might even get rid of the shortest silences

  • Try your settings on a small portion of your sequence before processing hours of footage

  • Don't work in the timeline while FireCut is working in the background

🙋‍♂️ Common questions, issues, watch-outs

  • "It cut out my entire sequence" --> The detection threshold was too high / the tightness was too tight. Try Advanced mode with an auto-detected threshold (and move it lower if the suggested value seems too high), and set the min silence duration at least 750ms.

  • "It slows down over time when processing long sequences" --> This happens with Original and Rapid algorithms because the number of clips on your timeline increases as cuts are made, slowing down Premiere Pro. If you're experiencing very long processing times (e.g. 30+ mins), then it's a sign that:

    • You are likely cutting more silences than you need (slowdown typically happens when you're cutting 200+ silences)

    • You should try Turbo mode

  • "I had an issue with Turbo mode" --> This feature is brand new, and we want to fix whatever caused the issue for you! Please use the "Share feedback" button inside FireCut / email support@firecut.ai

Last updated