# Video editing and smart talking-head cuts > Open the video editor on the right of a chat, have the AI cut talking-head footage and add subtitles and animation, edit the timeline, effects, and built-in animations by hand, and export in the background. - Source: https://getbeav.com/en/docs/workflows/video-editing Checked 2026-10-07. Applies to Beav desktop 2.8.13; the following section covers additions in 2.10.4. ## Script anchors and sound effects (2.10.4) Checked against source on 2026-11-02. A final script can use `@{name!}` point anchors and `@{name}…@{/name}` ranges. After captions align with the script, the AI can attach emphasis text, chapter cards, overlays, B-roll or sound effects to those words and realign them after a new voiceover or transcription. Manually moved or trimmed clips stop following. Their basic properties provide controls to resume or remove the attachment. Missing anchors, locked tracks and overlaps leave clips in place with a reported status. Saving a revised script tells the AI which paragraphs, voiceovers and on-camera sections need updating. Version 2.10.4 also ships an offline **SFX** tab with 67 effects, including 52 CC0 recordings usable commercially without credit. Search, preview and add effects at the playhead; adding inside a subsequence is not supported. Source and focused tests do not replace playback, sound-level and export checks after a real voiceover change. Beav's video editor opens in the right workspace of a chat, while you keep talking to the AI on the left. There is no separate editing page. ## Open the editor * In the right workspace, click **+** to open a new tab, choose **Video editor**, then choose a recent project, or choose **Create project** and import your media. * Or simply provide a video in the chat and describe how you want it edited. The AI creates or opens a project itself; you do not need to create one by hand first. Once a project is selected, your next message can direct the AI to edit. The AI can see the current project, the selected clips, the playhead position, and the I/O range. Use the top-left corner to go back to the project list and switch projects. Close the **Video editor** tab to leave editing and continue an ordinary chat. When you switch chats, each chat restores its own project. When the sidebar is narrow, the upper panel switches between **Preview**, **Media & tools**, and **Properties**, and the timeline stays visible. Drag the sidebar edge to make it wider. ![A Beav chat beside the video editor in the right workspace: the AI reports subtitle and title card progress on the left, the preview and video, text and audio tracks on the right](https://getbeav.com/screenshots/product-video.webp) *The editor opens in the right workspace. The AI reports each step in the chat, and you can preview and inspect the timeline next to it.* ## Smart talking-head cuts Provide a video in the chat and state your editing goal, for example "Clean up this talking-head clip and add subtitles." Lectures, courses, interviews, conversations, and screen recordings with narration are all handled with the talking-head workflow, as long as speech carries the video. The AI transcribes the video, identifies slips and pauses to remove, splits the speech into meaningful subtitles, and arranges the speaker shots, supplementary footage (B-roll), and explanatory animations. A broad request such as "edit this for me" includes full content editing, subtitles, B-roll, and animation by default. If you only want to trim one section, or explicitly don't want animation, just say so. * The plan from each analysis step is saved separately, and the AI reviews it before applying it to the timeline. You can also edit and save a plan in its plan dialog, then apply it directly without reanalyzing. * Picture-in-picture may start with a replaceable placeholder showing the needed visual and its position. Replace the media on the timeline, or ask the AI to fill it from project media, the asset library, or knowledge base assets. * If a request is interrupted, the saved project, transcript, and plans are still there. Ask the AI to continue; edits already applied are not applied again. ![A talking-head video loaded in the Beav editor, with the source clip on the timeline waiting to be cut](https://getbeav.com/screenshots/talking-head-editor-before-first-cut-2026-09-25.png) *The talking-head footage is loaded in the editor, waiting for the AI’s first cut.* ## Have the AI change specific content Select a clip and tell the AI your goal, for example "Make the selected subtitles white with a black outline and keep their position." Inside a group (subsequence), you can also say "Move the title in this group to the top." The AI can change text, position, size, volume, speed, and whether a clip is enabled. If no clip is selected, describe the specific time range or object. When you specify "only the I/O range," content outside the range is kept. AI changes go into the project's undo history, so you can undo them. ## Edit the timeline by hand * Click a clip to show a white selection box; drag it to move, and drag either end to trim. Click an empty area of a track to jump the playhead to that time, and drag across empty space to box-select. * The timeline and preview share the selection: clicking a subtitle, image, or video in the preview also selects the matching clip. * By default the timeline takes 45%–55% of the editing area's height. Drag the divider above it to adjust, and double-click the divider to reset. * Importing media only makes it available; it appears in the picture only after you place it on the timeline. If you can't see something, first check that the clip is on the timeline, then check the playback position and layer. * A layered animation appears on the main timeline as one sequence block. Double-click it to open its internal layers. ### Keyboard editing Click the timeline first so it has focus; the keys below only work there. This was checked against development source for an upcoming release; how it feels in the real desktop app hasn't been verified yet. | Key | What it does | | ---------------------------- | --------------------------------------------------------------------------------------------------------------------- | | J / K / L | Play backward / stop / play forward. Press L or J again to speed up to 2×, 4× and 8×; Shift+J / Shift+L play at 0.25× | | Alt+K | Play from the in point to the out point | | I / O, X | Set the in / out point; X marks the clip under the playhead | | `;` / `'` | Lift removes everything between in and out and leaves a gap; extract removes it and pulls later content back | | Q / W | Remove everything between the previous / next edit point and the playhead, and close the gap | | Alt+Q / Alt+W (or `[` / `]`) | Trim the selected clip's start / end to the playhead without closing the gap | | Shift+Q / Shift+W | Move the previous / next edit point to the playhead | | Cmd+Shift+V | Paste at the playhead as an insert, pushing later content along | | Shift+M / Cmd+Shift+M | Go to the next / previous marker | | Shift+I / Shift+O | Go to the in / out point | | Cmd+Shift+X | Clear in and out | | D | Select the clips under the playhead | Undo (Cmd+Z) and redo (Cmd+Shift+Z) work anywhere in the editor except text fields. ### Insert, overwrite and edit history Select a clip in the media library, set its in and out points in the source preview below it, then click **Insert** (or press `,`) or **Overwrite** (or press `.`). The clip goes to the timeline's in point, or to the playhead when no in/out is set. Insert pushes later content along; overwrite covers what is there. It lands on the track of the selected clip when that track fits, otherwise on the main track or the first audio track. With the timeline focused you can press `,` / `.` directly. The arrow next to the undo and redo buttons lists the edit history. Pick any step to go back to right after it, or **Original state** to go back to the start. ### Lower music while someone speaks Select the background music, then click **Auto-duck music** under **Audio > Audio processing** in the inspector. The editor adds volume keyframes to the music wherever speech is found in transcripts (or wherever a video with sound plays when there is no transcript), with a 0.8-second ramp before and after; the reduction is adjustable. It replaces the music's existing volume keyframes and can be undone; run it again after big edits. This was also checked against development source for an upcoming release. ### Trim mode, rate stretch and track targets With the timeline focused, press Shift+T to select the edit point nearest the playhead on each track; the preview shows the frames on both sides of the cut and shift counters. Ctrl+T cycles between ripple, roll and trim; Alt+← / Alt+→ move the point one frame (five with Shift), and E moves it to the playhead. J / L move it continuously and K stops and commits; Space plays from 3 seconds before the cut to 2 seconds after; Esc leaves trim mode. Each move is one undo step. Press Shift+R or click the gauge icon in the toolbar for rate stretch: dragging a clip edge changes its speed (0.25–4×) instead of its content, and keyframes and fades scale along. The small bar at the far left of each track header shows whether the track is targeted; click it to toggle. Cmd+0 / Cmd+9 toggle all picture / audio tracks. Q/W, lift/extract, D, X and insert/overwrite only act on targeted tracks. ### Saved versions The editor saves a version every 5 minutes (skipped when nothing changed; the newest 20 are kept). At the bottom of the menu next to undo and redo, choose **Save a version now**, or **Restore** a version: it opens as a new project, leaving the current project and its undo history unchanged. This was also checked against development source for an upcoming release. ## Styled text, text animation and voice effects **Text > Styled text** on the left has 93 presets, filtered by Emphasis, Warp & FX, Variety, Titles, Emotion, Promo, Labels, Subtitles and Sets. Card thumbnails use the same text renderer as export, so what you see is what you get. Click a card to add it at the playhead. With text selected, the inspector also offers: * **3D block text**: under **Advanced text effects**, switch the shadow style to **3D block**. The text gets a solid extrusion along the shadow offset; change the shadow X/Y offsets to set its direction and depth. * **Text effects**: under **Advanced text effects**, choose Hollow, Neon, Glitch, Chrome, Stripes or Offset outline. * **Warp**: in the same place, bend text into Arc, Wave, Perspective, Skew, Bulge, Pinch, Stretch, Stagger or Stairs, then set **Warp amount**. Negative amounts reverse the shape, so an arch becomes a smile. Effects and warps combine with strokes, gradients, 3D block text and text animation. * **Text animation**: in and out animations are Fade, Zoom, Bounce, Slide up, Shake, Slam, Elastic grow, Drop in, From left, From right, From top, Spin and Flicker in; every out animation also fades. Loops are Pulse, Float, Shake, Heartbeat, Swing, Hop, Blink and Glow. Select a video or audio clip with sound and use **Audio > Voice effect** to switch between Voice boost, Broadcast, Phone, Walkie-talkie, Old radio, Megaphone, Echo, Hall, Deep and Bright. A voice effect applies to the whole media file and is processed together with noise reduction, EQ and compression. It can be undone; choose None to remove it. You can also ask the AI, for example "change these captions to the variety 3D style and slam them in", "make the title an arched neon word" or "add voice boost to this talking-head clip". This was checked against development source for an upcoming release; desktop behaviour is still to be verified. ## Built-in animations **Animation** in the editor's top-left corner opens **Narration overlays** by default, with 9 presets: **Key phrase**, **Speaker intro**, **Q & A**, **Myth vs fact**, **Big number**, **Chapter title**, **Three beats**, **Opposing contrast**, and **Step highlight**. Click a preset to change its text, theme color, and duration (3–30 seconds, 5 seconds by default), then click **Add to timeline** to add a layered clip at the current playback position. **Graphic accents** offers 28 symbol animations, including arrows, circles, hearts, and sparkles. Click one to add a 3-second clip. All built-in animations work offline and do not need an internet connection or a model. ## Effects and camera motion Select a video, image, or shape clip and open **Effects**: * **Effects** are grouped into **Basic**, **Lens**, **Distort**, **Retro**, and **Split & frame**. Hover to preview and click to add. There are also combinations such as **Interview focus**, **Soft portrait**, **Film projector**, **Retro pixel screen**, **Dream prism**, **Comic impact**, **Diptych**, and **Contact sheet**. Effects appear on a separate purple **Effects** track; drag the clip or its edges to adjust where it applies. * **Camera motion** includes **Push in and return**, **Two-step push in**, **Gentle breathing**, and **Tilt and settle**. The rhythm scales with the clip's duration, and you can keep editing the keyframes. All effects are computed on your computer and do not need an internet connection. **Soft portrait** processes the whole frame; it is not face retouching. Camera motion scales around the clip's center and does not track people automatically. ## Explanatory animations made by the AI When you need to show relationships, comparisons, processes, or mechanisms, the AI can make motion graphics with Remotion, render them into real videos, import them into the editor, and place them on the timeline to match the narration. It can also make a standalone explainer animation and deliver both the source code and the finished video. This requires installing the Remotion plugin under **Tools & Skills → Plugins** first, plus a local runtime such as Node. See [Skills, plugins, and MCP](https://getbeav.com/en/docs/agent/skills-and-plugins). After import, you can trim and move the animation video; to change the graphics inside it, ask the AI to modify the source code and render again. ![A science explainer editing project with the title card, subtitles and narration on the timeline](https://getbeav.com/screenshots/science-explainer-video-editing-project-2026-09-26.png) *A science explainer project: the title card, subtitles and narration track share one timeline.* ## Multilingual text rendering (development builds) The development update dated 2026-10-16 accepts accented Italian, French and Latin text, as well as Arabic and Thai. Arabic keeps its writing direction and connected letters, Thai wraps at word boundaries, and previews and exports show the same text. Noto Sans, Noto Sans Arabic and Noto Sans Thai are included; other text uses fonts installed on the computer. Missing characters prompt you to install a suitable font. Not every Unicode character or colored emoji has an available glyph, and font appearance may vary by computer. This update has not been released; desktop interaction testing is still pending. ## Built-in fonts and caption size (development builds) From the development update dated 2026-10-27, captions, flower text and preview thumbnails use open-source fonts bundled with the app, so they look the same on Mac, Windows and Linux, and bold or black weights no longer mix in other system fonts. The **Font** setting offers Sans (Noto Sans SC, default), Serif (Noto Serif SC), Kai (LXGW WenKai), Smiley Sans, KuaiLe, HuangYou and the Latin title font Montserrat. Traditional Chinese, Japanese, Korean, Arabic, Thai, Hebrew and Devanagari text automatically uses the bundled Noto fonts. The Chinese, Japanese and Korean fonts include common characters only; rare characters fall back to system fonts. Kai, Smiley Sans, KuaiLe and HuangYou have a single weight. New captions default to about 7% of the frame's short side (about 75 px on 1080-wide vertical video), talking-head style templates scale their captions the same way, and captions in existing projects are not changed. This update has not been released; desktop interaction testing is still pending. ## Built-in sound effects (2.10.4) From version 2.10.4, the editor's left panel has an **SFX** tab with 67 short sound effects for social videos: whooshes, pops and clicks, alert dings, right/wrong cues, comedy boings, impacts, glitches, heartbeats, a camera shutter, typing and applause. 52 of them are real recordings released under CC0 (public domain), so they can be used commercially without credit; the rest are synthesized by the app. They ship with the app and work offline. They are not added to your library, and no background music is bundled. Hover over an effect to hear it and click to add it at the playhead on a separate SFX track. For sounds whose peak is in the middle, such as whooshes, the clip is shifted so the peak lands on the playhead. Added effects are ordinary audio clips you can move, adjust, delete or undo. They can't be added inside a nested sequence; go back to the main timeline. Check real desktop playback and export sound levels with your material. ## Export Ask the AI to export, or export from the editor. Export runs in the background, so you can keep editing after submitting it. Export uses a snapshot of the project at submission time; later changes need another export. If you only specify a file name, the file is saved in the project's exports folder, and duplicate names are numbered automatically. When the export finishes, use **Share to phone** in the video preview to get it by scanning a code. ## Report an editing problem Click **Report a problem** next to the AI button in the top-right corner of the editor's title bar, describe the problem, and submit. We recommend keeping **Attach redacted error logs** checked, as it is by default. ## Precise trimming and project versions (development builds) Checked 2026-10-09 against desktop 2.8.14 development source. Desktop shortcuts, dragging, trim preview, and version restoration have not been manually verified. * **Trim mode:** Shift+T selects the nearest edit point; Ctrl+T switches ripple, roll, and trim. Alt+arrow adjusts one frame, or five with Shift. E aligns to the playhead; J/L trim dynamically, K commits, Space previews, and Esc exits. * **Rate stretch:** Shift+R selects the tool. Drag a clip edge to change duration and speed (0.25–4×). * **Track targets:** Click the small bar in a track header to target it. Cmd+0 / Cmd+9 toggle video / audio targets. Three-point edits, lift, extract, and related commands use targeted tracks. * **Project versions:** Changed projects are saved every five minutes, retaining twenty versions. The undo/redo menu offers immediate saving and restoration. Restoring creates an independent project copy and preserves the original. Failed saves do not report success; automatic saves retry at the next interval. ## Text effects and talking-head styles (2.10.0) Checked 2026-10-23 against the 2.10.0 release candidate, bundled guide, and source. Real-account generation, cross-platform interaction, and exported results still require runtime validation. The editor includes decorative text, 3D text, text animation and warping, video effects, and sound effects. The **Style** tab offers eight talking-head presets with intensity controls for captions, grading, music ducking, and progress bars. Applying a preset is one undo step. Direct application does not choose emphasis text for you; ask the AI for content-aware styling when needed. Play the full edit before export and check caption safe areas, sound levels, sync, and cuts.