Plain-Language Editing Prompts That Actually Work

Name the action and the one detail that matters

A useful prompt is a short instruction, not a paragraph about your project. The editor has to map words onto a tool: trim, merge, split, compress, convert, transcribe, or a specific image action. “Make this better for Instagram” does not say whether you want a crop, a cut, a compress, or a caption. “Trim video from 0:12 to 0:48” does.

Put the parameter that would be wrong if guessed in the same sentence. Times for a trim. A target format for a convert. A language for transcription. A transition name when you care which one. Leave the rest to the buttons when a picker exists — trim handles, transition grids, and voice lists are more reliable than typing a novel.

Prompts for video

Use clock times, not vague words like “the middle.” Examples that map cleanly: “trim video from 0:00 to 0:30”, “trim from 1:05 to 2:10”, “cut from 00:45 to 02:10”. For a hard join of two files you already added: “merge these videos”. For a soft join: “merge videos with crossfade transition”. For length without a story edit: “split video every 30 seconds”.

Size and delivery prompts are separate jobs. “Compress video” and “change video resolution to 720p” do not replace a trim. Run them when the file is too heavy to send, after you know the cut is right. “Mute video” is the right prompt when the picture stays and the sound goes. “Extract audio from this video” is the right prompt when you only wanted the soundtrack.

Add the second file before a merge or a music bed. A prompt cannot attach a clip you have not selected.

• Trim and cut: include start and end.
• Merge: add every clip first, then name the transition only if you want one.
• Compress or resize after the creative cut, not instead of it.

Prompts for audio and images

Audio follows the same pattern. “Trim from 0:00 to 0:45”, “merge these audios”, “convert to mp3”, “remove background noise”. Transcription should name the language when you know it: “transcribe this audio in English”. Text-to-speech is easier from the voice picker than from a sentence that tries to describe a voice in adjectives.

Images: “convert to webp”, “convert to jpg”, “convert to png”, “compress image”, “change aspect ratio to 1:1”. Background removal and image generation are quota tools with their own panels. “Turn these images into a video” belongs to the slideshow flow, where you also pick music and transitions, not to a one-line crop.

If the file type and the prompt disagree — a crop prompt on an MP3, a transcribe prompt on a PNG — fix the file selection instead of rephrasing. The words are fine. The input is not.

What to do when the prompt is ignored

Read the on-screen error before you rewrite the sentence five times. The usual causes are a file over the comfortable browser size, a missing second input, a daily AI limit, or timestamps that run past the end of the clip. Compress or shorten first when the error is about memory. Check the profile quotas when the error is about AI.

Interactive controls exist because some details are awkward in prose. A crossfade duration, a exact crop rectangle, and a TTS voice are easier to set with a control than with adjectives. Use the prompt to choose the tool, then use the control to set the number.

Keep a personal list of the ten prompts you actually use. You do not need a different sentence for every project. You need one precise sentence per job, a file that fits, and a preview before you download.