Writing ClipAnything prompts

How to describe the moments you want so ClipAnything returns clips you would actually post.

What ClipAnything is for

OpusClip has two clip models. ClipBasic works from what is said, so it suits talking-head content. ClipAnything also reads visual, audio and emotional cues, which means you can ask in plain English for a scene, an action, a person, an event or a reaction, even when nobody says the words out loud. The prompt is optional, but a good one turns a random batch of clips into the ones you actually wanted.

Build the prompt from parts

A reliable prompt names the details that matter to you:

- Subject: the person, animal, object or place. - Visual: something that makes it recognisable, like a shirt colour. - Action: what happens. - Sentiment: the emotion, such as surprise, anger or laughter. - Sound: an audible cue like cheering or applause.

You do not need all five every time. "The guest in the red hoodie laughs after the host's joke" already covers subject, visual, action and sentiment. For compilations, ask for every instance, for example "all moments where the audience applauds".

What to leave out

The prompt is for finding moments, not for editing. Leave out clip length (there is a separate Clip Length option), aspect ratio, captions, transitions, effects and emojis; those belong in the settings, brand template or editor. Skip requests that describe no content, such as "make a video for Instagram". Describe what you want rather than what to exclude, and write the prompt in English.

Where to prompt and how to iterate

You can enter a prompt on the submission page before processing, or on the result page once the project is done. The second option is the useful one: if the first batch misses, rewrite the prompt and run it again instead of uploading again. The result page also shows suggested prompts you can click.

When it finds nothing

ClipAnything struggles with silent footage and static scenes, because there is little sound or movement to work with. If results are off, give the thumbs-down on bad clips and make the prompt more specific about who, what, and what it sounds like.