Turn long videos into short clips with captions, on your computer

To turn long videos into short clips, you need the words, so you can find the moments worth keeping; a quick way to cut; captions people can read without sound; and a frame that keeps the speaker in view on a phone. Barside's Clips editor does all of it on your own computer. Add a video to a note or record your screen, choose Make clips, and cut by editing the words. Barside adds animated captions, reframes the video to 9:16, 1:1 or 16:9 around the speaker, and renders an MP4 back into the note.

Your video is never uploaded. If you want AI to find the best moments, your own Claude or ChatGPT can do that part through Barside's AI connections, reading only the transcript.

How to turn long videos into short clips in Barside

1. Put the video in a note. Drag a video file into a note, paste it, use the media button, or record your screen. The Clips tab in the sidebar lists every video in your notes: Add video puts each file in its own new note, and Record starts a new recording. MP4, WebM, MOV and M4V files work.
2. Choose Make clips under the video, or on its card in the Clips tab. The editor opens over the window.
3. Install the video tools, once. The first time, Barside offers to download FFmpeg 8.1.3 (LGPL): about 70 MB on Windows, 50 MB on Linux and 35 MB in Barside for Mac. Choose Install video tools.
4. Let Barside write down the words. The local speech engine transcribes the video, and Barside makes a first clip: the whole video if it is 3 minutes or shorter, otherwise its first minute. If the speech engine is not installed yet, install it once in Assistant settings (about 150 MB), then choose Get the words.
5. Edit the clip in four tabs: Words, Captions, Format and B-roll.
6. Choose Make clip. Barside renders an MP4 and adds it to the note right after the source video. Save as MP4 copies it anywhere, ready to post.

Done closes the editor. Your clips and edits are saved with the video, so you can come back later, change a caption and choose Make again. The search bar finds your videos by name too, and opens them straight in the editor.

Edit the video by editing its words

The Words tab shows the transcript word by word, and the word being spoken lights up as the video plays.

• Drag across words to select them; Shift-click extends a selection. Double-click a word to jump to it.
• Cut removes the selected words from the clip, and Delete does the same. Cut words stay in the list, crossed out. Select them and choose Keep to bring them back.
• New clip makes a separate clip from the selection, so one long video can hold many short clips, up to 50.
• B-roll opens the B-roll tab to place a photo or video over the selected words.
• Remove filler words cuts "um", "uh", "erm", "hmm" and similar sounds in one step.
• Shorten pauses trims every pause longer than 0.7 seconds down to 0.25 seconds.
• Whole video shows all the words, not only those near the clip.

Under the preview, the timeline shows the clip's range in the whole video: drag its edges to start earlier or end later. A second bar shows the clip itself, with its cuts and B-roll, and its length after cuts. Playback skips what you cut, so you hear the result as you edit. Space plays or pauses, and Escape clears a selection.

The speech engine writes down the words in the language set in Assistant settings. English is included; automatic detection and 34 other languages come with a one-time 141 MB download.

Captions that follow the voice

The Captions tab styles the words on screen.

• Style: Bold and Boxed show the whole caption and make the spoken word pop in the highlight color. Karaoke fills each word with the color as it is spoken. Clean shows plain words, and None turns captions off.
• Position: Top, Middle or Bottom.
• Size: S, M or L.
• Highlight color: yellow, green, blue, pink or white (every style but Clean).
• Words per caption: 1 to 5.
• Capital letters: on or off.

A new clip starts with Bold captions at the bottom, three words at a time, in capitals, highlighted in yellow. Captions are burned into the MP4, so they look the same wherever you post it.

Reframe to vertical, square or wide

The Format tab sets the clip's Shape:

Framing decides what stays in view when the shape changes:

• Follow the speaker, the default, finds faces on your computer with MediaPipe BlazeFace and moves the frame to follow the biggest face. The editor shows its progress while it looks.
• Center keeps the middle of the picture.
• Whole picture, blurred sides fits the entire video and fills the rest with a blurred copy. It suits screen recordings, where cropping would cut off part of the screen.

Clips render as H.264 MP4 files with AAC sound, at 30 frames per second. On Windows, Barside uses the system's own H.264 encoder and falls back to OpenH264 if that fails. On Linux, it uses OpenH264, and Barside for Mac uses Apple's VideoToolbox encoder.

Add B-roll from your own photos and videos

B-roll shows a picture or another video over the speaker while the sound continues. Select words first to place it exactly on them; otherwise it starts at the playhead for 3 seconds. Add B-roll lists the photos and videos already in your notes, so the clip uses your own media, not stock footage. Remove an item with its trash button. A clip can hold up to 60 B-roll items.

Let your AI find the best moments

The editor works without any AI. When you want help choosing, your own AI app can do the AI parts: read the transcript, pick the moments worth a clip, add B-roll where it fits and make the clips. This runs through Barside's AI connections, so it works with Claude Desktop, ChatGPT, Cursor and other MCP apps.

1. Open Settings → Use Barside in your AI app → More apps and options. Pick a connection, such as the one Add to Claude Desktop made, or create one with New connection.
2. In its Permissions tab, share the workspace that holds the video. Turn on Allow adding and updating notes, bookmarks, tasks and reminders, then Allow making video clips.
3. In the clips editor, choose Ask your AI and copy one of the ready requests into your AI app, like: find the 3 best moments (30 to 60 seconds each) in my video and make each one a 9:16 clip with bold captions.

The AI reads timed sentences from the transcript, not the video: it never receives the video files. Clips it makes appear in the editor with its reason for choosing each one, and it can give each clip a score. The tools your AI can use are listed on the AI connections page.

Who uses Clips

• Podcasters who record an hour and want four short highlights with captions for the week.
• Course creators who turn a lesson recording into short previews for social media.
• Founders and product teams who cut a demo into a vertical teaser and a wide version for the website.
• Community owners who post highlights of a live session for the members of their Space.
• Teams who record a video meeting, which saves the recording in the meeting note, and clip the decision for people who missed it.
• Speakers and interviewers who want the filler words gone before they post.

If this is most of your work, switch to Creator Mode. It keeps Recordings, Clips, drawings, your notes and Spaces in the sidebar, and opens on the Recordings tab, where each video has Make clips. See app modes.

What it costs and what it needs

Clips is free: no minutes to buy, no credits and no watermark on the MP4. It needs:

• The video tools: FFmpeg 8.1.3, a one-time download from Barside's own download storage, checked against a fixed SHA-256 before it is used. About 70 MB on Windows, 50 MB on Linux and 35 MB on a Mac.
• The speech engine: Whisper, installed once from Assistant settings. It is about 150 MB, plus 141 MB for languages other than English, and needs no AI key.
• An AI app, only if you want AI to pick the moments. Its company bills you as usual.

If a video cannot play in the editor's preview, clips from it still render. Clips runs on Windows 10 and 11, on 64-bit Linux and on Mac.

Privacy

The clips editor cuts, captions, reframes and renders on your computer. The speech engine and the face model run on your computer too. Each video's clip projects, words, pauses and face positions are saved next to the video in your media folder, so your backups carry them. Nothing is uploaded while you make clips. A finished clip leaves your computer only when you share it with a private link, or with your library if you turn on Barside Plus cloud backup or cloud sync. See Clips and your data.

If you allow video clips for an AI connection, that AI app can read the transcripts of videos in the notes you share with it, list your photos and videos there for B-roll, and make clips from them. It never receives the video files.

How it compares

Web clip makers such as OpusClip work online: you drop in a video link or upload the file, and their service makes the clips. Barside keeps the video on your computer, lets you control every cut by its words, and brings in your own AI only when you want it. Read the OpusClip alternative for a side-by-side comparison, or see the best AI clip makers.

To get a written transcript of a call rather than clips, use call transcription.

Works well with