Video Editing Prompts for Cuts, Text and Audio

Specify the action, target and time range

Start with an editing verb: add, remove, mute, increase, resize or split. Follow it with the target, such as the selected clip or the interval from 00:08 to 00:12. Add a constraint when it matters: keep the footage, leave the audio unchanged or preserve the existing order. The agent can then map the instruction to actual project operations instead of guessing a style.

Use times from the current project. Removing five seconds near the beginning shifts everything afterward five seconds earlier. A timestamp copied from the original recording may identify a different moment once you have made cuts. Move the playhead to the intended point and read its position before writing a follow-up.

Mute the audio from 00:08 to 00:12. Keep all the footage and leave the sound outside that range unchanged.

Give exact words and timing for video text

Put the required text in quotation marks and state whether it belongs at the top, center or bottom. Include a start and end, rather than a vague instruction to show it during the intro. A short title and a dark text background are easier to read over changing footage than several lines of small text.

Review the result while the video is moving. A title that looks clear over the first frame may overlap a face or disappear against a bright background a second later. Check the exported file too, since it contains the rendered layer rather than a separate text box.

Add “Three ways to save time” at the bottom from 00:00 to 00:03. Use white text with a dark background. Keep the clips and audio unchanged.

Ask for cuts without losing the meaning

For a known unwanted section, provide its current start and end. For several removals, describe each range clearly and review every join. A technically correct cut can still remove the start of a word, interrupt a gesture or leave an abrupt change in the speaker’s position.

Pause requests can use local audio silence analysis. That evidence identifies quiet ranges rather than the meaning of the words around them. Do not assume the agent understands every sentence in a long recording or can find an event that is absent from its sampled frames. Give your own timestamps for important decisions.

Remove the section from 00:14 to 00:17 and join the remaining footage. Do not change the playback speed.

Make audio and framing instructions measurable

Specify a volume percentage or a time range to mute. A request to improve audio can mean several incompatible things: raise the voice, remove a sound, lower everything or change the speed. Increasing gain also raises background noise and cannot restore information lost in a clipped recording.

For resizing, name the aspect ratio and whether you want to keep the whole frame or crop to fill it. A centered 9:16 crop may remove objects near either side of a landscape recording. Watch the full clip to confirm that moving subjects remain visible.

Make the project 9:16. Keep the whole source frame visible rather than cropping the sides.

Choose useful context instead of adding more

The selected clip, existing timeline and prompt supply important context. Include frames only when visual evidence helps. Sparse image samples are not a full video transcript, and a brief action between samples can be missed. A clear timestamp often matters more than a long description of the scene.

Quick and Standard use OpenAI language models; Advanced uses an Anthropic model. Every model is limited to the editor’s supported operations. A more expensive choice cannot enable object removal, automatic transcription or generated footage when those tools are absent. Check the maximum credit reservation before the request.

Review one change before asking for the next

Play the affected section and compare it with the original project state using undo when necessary. Check picture, timing and sound together. If the wrong clip changed, undo, select the intended clip and narrow the instruction rather than layering a correction onto an uncertain result.

Manual adjustments and browser exports do not spend AI credits. Use them for a small timing correction when the agent’s first pass is close. Keep the original file and download a project backup so you can recover the edit instructions if browser storage is cleared.

Frequently asked questions

Related video editing guides

• Trim Video Online: Cut the Beginning, End or Middle
• Add Text to Video Online with Exact Timing
• Remove Silence from Video and Make Jump Cuts