I spent 2 October making Cutwise more useful once the automatic editing suggestions have done their job. The work started with a waveform I could cut directly and grew into separate video and audio tracks, editable captions, extra media layers and animated callouts.

In the first Cutwise post, I described getting a recording through transcription, review and export, then packaging the Mac application. This time the focus was the part between accepting an edit and being happy with the result: adjusting a boundary, keeping the picture while silencing some audio, placing an image or correcting a caption.

Those are small decisions, but they need proper controls. As I added them, preview performance, undo and the less glamorous details of finding and removing things became just as important as the new features.

Starting with the waveform

The first change was a waveform editor with range selection and zoom. It gives the audio a visible shape, so I can select a pause or unwanted sound without having to express every decision as a pair of timestamps.

I kept two actions distinct. Cutting removes the selected time from the edited video and closes the gap. Silencing keeps the picture and duration while muting the selected audio. That distinction matters in a screen recording: an unwanted noise does not necessarily mean the demonstration on screen should disappear too.

The edited playback follows those decisions, so the result can be heard before export. The original recording remains unchanged, and transcript timestamps still refer to that original source.

I then added independent video and audio editing. Clips can be split on the selected tracks, and inserting an image or another video becomes an explicit timeline operation. A split creates a boundary; removing a section is a separate decision.

That gave Cutwise a more direct editing surface while keeping the existing rule that every change is a reversible decision about the media.

Giving captions their own place

Captions already existed, but generated subtitles still need attention. A recogniser can get a word wrong, and a technically valid caption can appear at an awkward moment.

I added an editable caption track above the video. Caption blocks can be dragged to a new position, trimmed at either edge or opened to change their text and precise timing. Style changes appear in the video preview, and the edited track supplies both the rendered captions and subtitle exports.

The important part was connecting those changes to the rest of the timeline. Undoing a media edit needs to restore the caption timing affected by it too.

I brought video, audio and caption edits into one undo history, with buttons that describe the action they will reverse. Splits, inserted clips, cuts, silences and transitions can all be stepped backwards and forwards through the same controls. The history keeps the last 50 edits in the current session; it resets when a project is opened, the source is replaced or the recording is transcribed again.

Autosaving and undo solve different problems. The project should remember the current edit, while the editor should still let me change my mind about how I got there.

Adding images and sound around the recording

The next step was extra media tracks. Cutwise now supports additional video tracks for images and video overlays, plus audio tracks for music or other sound. Imported assets are copied into the project so these additions travel with its files.

Images can be positioned directly over a paused preview. Dragging moves the image, and corner handles resize it while preserving its proportions. There are precise controls for timing and placement as well, but being able to put something where it belongs by looking at the actual picture makes the operation much easier to judge.

Track order controls which visual layer appears on top. Tracks can be rearranged, hidden or muted, and several can be selected for a single split at the playhead. Background music starts at a reduced volume, while imported video overlays start silent.

I also kept the current limits explicit. Extra clips stay at their positions in the edited timeline when the main video changes, and the export ends with the main video. Clips cannot overlap on the same extra track, so simultaneous overlays belong on separate tracks. There is room for up to 16 additional tracks.

One smaller follow-up was making inserted images easier to remove. I added a visible Delete image action and the same option in the clip's context menu. Removing an inserted image closes its gap on both main tracks and can be undone. Adding media is only half a usable feature if removing it requires hunting through the interface.

Making callouts editable on the timeline

With the extra tracks in place, I added manually editable animated callouts to them.

The callout editor offers 12 designs, six accent colours and separate entrance and exit animations. Text, position, size, opacity and timing can be changed, with motion visible in the editor. A callout can fade or slide in and leave with a different animation; very short clips automatically shorten those animation times to fit.

Once added, a callout behaves like the other visual clips. It can be moved, extended, split, hidden, reordered or removed, and double-clicking opens it for further editing. Its graphics are stored inside the project, and the animation is included in playback and exports.

This gives the finishing stage a useful manual path. If a particular instruction needs a label, I can write it, place it and adjust how long it stays visible without running another analysis.

Keeping previews useful as the timeline grows

More controls also mean more reasons for a preview to change. Rebuilding the base video for every adjustment would make simple visual decisions unnecessarily slow.

I changed preview generation to reuse cached four-second sections and render the sections affected by an edit. Images and video overlays are displayed live over that base preview, so changing their position, size, opacity or stacking order does not rebuild the underlying video. Audio changes still update the cached mix.

Media verification received similar treatment: unchanged files can reuse their verification while editing, while exports fully verify the source and imported assets. The first preview still needs to build its cache, so this does not make preparation disappear.

The difficult boundary is keeping those cached sections consistent with the exported result. I added regression coverage for edits around section joins, seeking late in a recording, fractional frame rates and transitions that cross a section boundary. The tests compare preview frames with a full render and check that cancellation cannot leave a partial preview marked as ready.

That work matters because a responsive preview is only useful if it remains a trustworthy way to judge the edit.

Showing effects before choosing them

I expanded the video-track transition picker to 49 choices grouped into fades, wipes, slides, shapes and effects. The names alone are not always enough to explain what a transition will do, so the picker includes animated samples rendered with the actual effects.

These timeline transitions preserve the audio, captions and overall sequence length. The existing cleanup suggestions retain their more restrained options for joins around speech. A decorative transition between clips and a safe cut through a spoken explanation need different treatment.

I also added full-resolution PNG frame extraction. Extract frame pauses at the selected position and saves the composed result with cuts, inserted media, transitions and visible captions applied. It uses the source resolution with square pixels, rather than saving a screenshot of the smaller preview window.

That makes the finished composition useful outside the video too, without requiring a separate export just to capture one frame.

Making the controls easier to find

The growing interface needed more explanation. I added contextual help throughout the application, with a saved preference for whether it is shown, then corrected its appearance in dark mode. Links from editing panels now open the relevant Settings section, and runtime feedback distinguishes a development build without bundled tools from an incomplete packaged application.

The transcript gained two useful changes. It can be hidden to give the video more space without losing playback or the transcript's scroll position. It can also be searched for words and phrases, with highlighted matches and previous and next controls.

Searching deliberately leaves playback alone. Moving through results lets me browse the text; clicking a word still seeks the recording. That keeps finding a passage separate from interrupting whatever is currently playing.

I also expanded support diagnostics with sanitised failure details, including tool exit codes, recognised technical errors and package stack locations. The exported archive still excludes recordings, transcript text, source paths and credentials, and nothing is uploaded automatically. A failure report should help explain which operation broke without carrying the recording's contents with it.

Where the editing work ended up

Cutwise now has a much more capable place to finish an edit: direct waveform controls, independent media tracks, movable captions, extra layers, animated callouts and a shared undo history. The preview work supports those controls, and the smaller interface changes make them easier to discover.

This is progress in the development build. The documented signed Mac candidate remains marked INTERNAL ONLY, and the separate Windows, clean-Mac and distribution checks from the previous post are still outstanding. Adding editing features has not completed those checks.