> For the complete documentation index, see [llms.txt](https://learn.firecut.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://learn.firecut.ai/features/magic-cut.md).

# Magic cut

### ❓ What is it?

Magic cut turns a raw, unedited talking-head recording into a first edit.

If you record talking-head or "self-tape" style videos (several takes, false starts, stumbles and chat between takes), you usually face lots of manual work before the video is watchable: cutting the bad takes, tightening pauses, adding captions, zooms and chapters.

Magic cut does that first pass for you. It transcribes your footage, works out which parts are false starts, repeated takes, filler words and chat between takes, removes them, and builds a new sequence from what's left. Depending on your video, it also adds zooms, captions, titles or chapters.

You give FireCut one thing:

* One **video clip or sequence** with speech in it

FireCut then makes a new sequence, and your original clip or sequence isn't changed.

### 🤔 How does it work?

The feature works in three stages, all automatic after you click **Go**.

#### Step 1: Transcribe your footage

**Pick a video clip or sequence from your project** (bins expand and collapse; a small icon shows whether each item is a video file or a sequence) and click **Go**. FireCut transcribes it and saves the transcript, so the next run on the same source is quicker.

If you've changed the source (e.g. you moved some things around in a sequence that was previously transcribed), **you can force a new transcription**: click the three-dots button on the right of **Go** and choose **Re-transcribe and Go**.

#### Step 2: Cut the timeline

FireCut works out what to remove: false starts, repeated takes, filler words and talk between takes (like "let me try that again"). When you've recorded a line several times, it keeps one complete take and removes the others.

It then builds a new sequence containing only the kept parts, called "FireCut Magic - " followed by your clip's name, in the Script bin inside the FireCut bin:

* Each cut is trimmed against the audio, so breaths and trailing sounds are removed too
* Long pauses inside a kept section are cut too
* A short audio crossfade is added at every cut, so edits don't pop or click

When it's done, FireCut tells you how much it removed, for example "Footage cut down from 12:40 to 8:05 (4:35 removed)".

#### Step 3: Add the extras

Magic cut then adds extras, each on its own track, so the result stays easy to edit:

* **Zooms:** a zoom at the cuts, using Adjustment Layers on V2
* **Intro punch-out:** a quick blur and zoom-out at the very start, on V3
* **Captions:** on V4
* **Titles:** key phrases as titles, on V5
* **Chapters:** a chapter clip at the start of each chapter

On **Auto**, FireCut looks at your video and picks the extras that suit it. For example, vertical videos don't tend to get chapters, and horizontal videos don't tend to get captions and titles.

You can set each one to **On** or **Off** under **Advanced options** (see [Settings](#settings) below). To drop one after the run, disable or delete its track.

{% embed url="<https://www.youtube.com/watch?v=OskbuFBvRfg>" %}

### ⚙️ Settings

There are 2 main settings for Magic cut:

1. **Source clip or sequence:** The clip or sequence to edit, picked from your project. Expand bins to find it, or type in the search box to filter the list. You can select one clip or sequence at a time. A captions icon shows the items FireCut has already transcribed
2. **Advanced options:** Click this row to open more settings. When it's closed, it shows your audio language and anything you've changed:
   * **Audio language:** The language of the speech. The default, **Auto-detect**, lets FireCut work out the language itself. Set it yourself if the transcript comes out in the wrong language
   * **Target duration:** How long you'd like the result to be: **Auto** (the default, no target), **1:00**, **3:00**, **10:00**, or **Custom…** to type your own length (m:ss). The target must be at least 30 sec and shorter than your source, and options as long as your source or longer are greyed out. FireCut aims close to the target rather than hitting it exactly. Anything removed only to reach the target shows as **Trimmed for length** when you review the cuts
   * **Guidance script:** Click **Choose .txt…** (or drag a `.txt` file onto it) to give FireCut the script your video should follow. FireCut then keeps the take closest to your script and removes anything that isn't in it, even well-delivered sections. The file can be up to 2 MB, and no more than twice as long as the transcript. It's only used for the next run, so choose it again each time
   * **Review cuts before applying:** Off by default. Turn it on to check and change the cuts before the new sequence is built (see [Reviewing the cuts](#reviewing-the-cuts) below)
   * **Pacing:** How much room is left around what's kept. **Balanced** is the default and removes long pauses inside the parts it keeps. **Fast** leaves very little room around each kept part and removes shorter pauses too. **Slow** leaves the most room and only removes the longest pauses
   * **Zooms**, **Intro punch-out**, **Chapters**, **Captions** and **Titles:** **Auto** (the default) lets the shape and length of your video decide (see [How does it work?](#how-does-it-work) above). **On** always adds it, and **Off** never does
   * **Preset:** Next to **Captions** and **Titles**, this picks their look. **Default** is FireCut's standard Magic cut style, and you can also pick any of your caption presets or title themes, including your own

### ⚒️ Reviewing the cuts

By default, Magic cut builds the new sequence straight away. If you'd like to approve the cuts first, open **Advanced options** and turn on **Review cuts before applying**.

With review on, clicking **Go** runs the analysis and then stops at **Review cuts**, before anything changes in your project. It shows your full transcript: green text is kept, and each run of removed text is folded into a line such as "12 words removed". Click that line to see every removed part, struck through and coloured by why it was removed.

* **Adjust a section:** Drag the edge of green text to trim it. You can drag it back out again, but only as far as where that section first ended.
* **Keep something that was cut:** Click a removed part to keep it, and click it again to remove it. You can also right-click text you've removed and choose **Keep this section**.
* **Remove a section:** Right-click green text and choose **Don't keep this section**.
* **Split a section:** Right-click green text and choose **Split section**.
* **Jump to a kind of cut:** The buttons at the top (**False starts**, **Meta**, **Repeats**, **Fillers**, **Trimmed for length** and **Other**) jump to the next cut of that kind. **Meta** is talk about the recording itself, like "let me try that again".
* **Keep or cut a whole kind:** The small menu on each of those buttons has **Keep all** and **Cut all**.
* **Start again:** **Reset all** undoes all your changes. **Expand all** opens every removed part.
* **Create timeline:** Builds the new sequence from your edited selection.
* **Cancel:** Goes back to the Magic cut screen. Nothing is changed.

The top of the panel shows the length of your source, the new length and how much is removed. The final sequence can come out slightly shorter than this, because FireCut also trims the ends of words against the audio when it builds it.

### 💡 Tips for best outcomes

* **Use clean audio.** Mute background music and sound effects, and keep speech clear — the cleaner the transcript, the better the cut decisions.
* **Record with cuts in mind.** If you flub a line, just pause and start the line again — repeated takes are exactly what Magic cut is built to resolve. Long pauses between takes actually help, both for detection and for accurate trimming.
* **Iterate freely.** The transcript *and* the cut analysis are both cached, so re-running Magic cut on the same source is fast and consistent — experiment with review mode without waiting.
* **Treat the output as a strong first cut.** It's designed to be edited: every enhancement is a normal clip on its own track, so you can tweak or remove any piece in Premiere.

### 🙋‍♂️ Common questions, issues, watch-outs

* **"It cut a take I wanted" / "It kept the wrong take"** --> Magic cut keeps one complete take of each line, and it might not be the one you'd pick. To choose a different one, turn on **Review cuts before applying** and click **Go** again (the analysis is saved, so it's quick). Open the removed text, click the take you want to keep, then right-click the take you don't want and choose **Don't keep this section**
* **"It says nothing was cut"** --> You'll see "No filler words, false starts or repeated takes were found in this clip, so nothing was cut." when there was almost nothing to remove, which usually means the recording was already clean. The zooms, captions and other extras are still added
* **"The result is much shorter than my target duration"** --> FireCut removes the mistakes and retakes first, and it can't make the video longer than the usable material that's left. When the result is well short of your target, you'll get a message saying how much usable material there was
* **"The cuts feel too tight" / "Words get clipped"** --> Set **Pacing** to **Slow** and click **Go** again
* **"I got captions, titles or chapters I didn't want (or didn't get ones I did)"** --> Set each one to **On** or **Off** under **Advanced options**. On **Auto**, captions and titles are only added to vertical videos, and chapters only to horizontal videos that are 7 minutes or longer after cutting
* **"The transcription language is wrong"** --> Set **Advanced options** > **Audio language** to the language of your audio and click **Go** again. FireCut transcribes the clip again in that language. To transcribe again with **Auto-detect**, click the three-dots button on the right of **Go** and choose **Re-transcribe and Go**
* **"A title covers words that are missing from the captions"** --> That's intentional: the caption clips under each title are disabled, not deleted. To reject a title, delete the title clip on V5 and re-enable the disabled caption clips under it on V4
* **"Will this change my original clip or sequence?"** --> No. Magic cut builds a separate new sequence ("FireCut Magic - " followed by your clip's name). Your source is untouched
* **"I ran it twice and got the same cut"** --> The cut analysis is saved, so running it again on the same source with the same **Target duration** and **Guidance script** gives the same cuts, and gives them quickly. Changing either one gets you a new analysis. To start completely fresh, click the three-dots button on the right of **Go** and choose **Re-transcribe and Go**
* **"Can I run it on multiple clips at once?"** --> Not yet: Magic cut works on one clip or sequence at a time. To edit a multi-clip shoot, put the clips in a sequence first and select that sequence as the source
* **"There's an extra copy of my source clip in the FireCut bin"** --> That's a side effect of how Premiere Pro imports the new sequence. It points to the same media file (it isn't a duplicate on disk), and the new sequence uses it, so leave it in place
* **"My video's waveform is now missing from the timeline"** --> This can happen if you used a sequence as the source rather than a video clip. To fix it, go to **Sequence** > **Render Audio**, which tells Premiere Pro to draw the audio waveform
* **"It says my source appears to be silent"** --> FireCut found almost no sound at the start of your source. Check that its audio tracks aren't muted or empty. If your video just starts with a long silence and the speech comes later, click **Proceed anyway**
* **"Magic cut request failed"** --> Please try again. If it keeps failing on a long recording, split the video into shorter parts and run Magic cut on each one
* **"The transcript has mistakes"** --> Because audio is transcribed automatically, some errors are expected, and Magic cut is built to cope with them. If the results are badly off, check that the audio is clean and **Audio language** is right. If it keeps happening, please share your audio with us at <support@firecut.ai>
