Playcut v2 is coming MCP · Actor v2 · & much more
Prompt library

Prompts sorted by the model that will read them

An AI prompt library sorted by the model that reads each prompt and by the job it does, with each model's reference-image syntax and prompt limits.

Updated · Playcut prompt library

Attach the same product photo to four video models and each expects a different name for it: [Image 1] for HappyHorse 1.1, a bare Image 1 for Wan 2.7 and <IMAGE_REF_0> for Gemini Omni Flash, which counts from zero. Seedance 2.0 wants the label written next to the subject it shows, and Nano Banana takes no tag at all. Copy a prompt across models and it can point at the wrong picture, or at none.

So this prompt library sorts prompts by AI model first and by job second. The builders below also cover models Playcut doesn't run, and say so.

How each model we run reads a prompt

Each row is a model in the Playcut studio: how a prompt names its reference images, and the rule most likely to change what the model reads. Tags match the Playcut developer docs unless a row says otherwise, and credit costs are on the Playcut models page. Each video and music row also links the prompt page written for that model.

Model Reference tag Also watch for
Nano Banana Pro and Nano Banana 2 No tag. Attach the most important image first. Up to 14 references, read in upload order.
Grok Imagine Image No tag. Describe the change in plain words. Two references at most.
Veo 3.1
Veo 3.1 prompts 		No tag in our docs. Describe each subject. Google caps prompts at 1,024 tokens; Playcut won't shorten a longer one.
Seedance 2.0
Seedance 2.0 prompts 		Image 1 , Image 2 next to the subject, as ByteDance writes it (Playcut's API docs write @Image1) An unnamed reference is effectively ignored. Faces in input images are rejected.
HappyHorse 1.1
HappyHorse prompts 		[Image 1] , [Image 2] Negative prompts are rejected. Text past 5,000 characters is silently cut.
Wan 2.7
Wan 2.7 prompts 		Image 1 , Image 2 (no brackets) The provider expands every prompt.
Grok Imagine (video)
Grok Imagine prompts 		No tag listed here: xAI's current docs and ours disagree on it. Name each subject in words. Playcut accepts up to 7 references on the classic model to steer look and style.
Gemini Omni Flash <IMAGE_REF_0> , <IMAGE_REF_1> (counts from zero) Negative prompts are rejected. A saying: "…" clause is spoken.
Lyria 3 (music)
Lyria 3 prompts 		No tag. Up to 10 images set the mood. Real artist, album and song names are blocked. Describe the style.

Playcut's image and video tools on the API and MCP cap prompts at 8,000 characters, and the music tool at 5,000. The studio has no built-in prompt enhancer; the AI Prompt Enhancer listed on this page is a separate browser tool. The rewriting you may notice comes from providers: Wan 2.7 expands prompts and Lyria 3 runs Google's own rewriter.

Browse the AI prompt library by model, job or method

Each page holds its own prompts or method, so this hub stays an index. Pick the model you use or the job in front of you.

Prompt builders, including models Playcut doesn't run

Each builder turns a short form into a prompt, in your browser. A builder for a model Playcut doesn't run says so on its card, and its output belongs in that model's own app. All of them are in the Playcut free tools directory.

For a model Playcut runs

For models Playcut doesn't run

For any model

Save and reuse prompts inside Playcut

Inside the Playcut app, the saved-prompt library keeps any prompt you want to run again, sorted into groups. You can also save a prompt straight from a generated asset's preview, so the wording behind a result you liked isn't lost.

To reuse one, type / at the start of an empty prompt box in the image, video, music or Act app. A picker opens and drops the saved prompt in, ready to edit for the model you picked. No link can pre-fill a prompt from this site, so copy it here and save it once in Playcut.

FAQ

Are the prompts in this library free to use?

Yes. The prompt pages and builders here are free to use in your browser, with no account. Running a prompt inside Playcut needs a paid plan, from $9/mo, and every plan includes a commercial license.

Can I paste a prompt written for one model into another?

Mostly, once you fix the reference tags and check the new model's limits. Seedance 2.0, HappyHorse 1.1, Wan 2.7 and Gemini Omni Flash each use a different tag, and Omni counts from zero. Nano Banana and Grok Imagine Image take none, so describe each subject in words. HappyHorse and Omni also reject negative prompts.

Does Playcut pick the model for each prompt?

No, you pick it. The image and video apps each have a model picker, and the image picker shows credit costs before you generate. Images default to Nano Banana Pro and video to Veo 3.1.

How do I use a prompt from this library in Playcut?

Copy it, open the matching app in the Playcut studio, pick the model the page names and attach any reference images it mentions. Then paste. The app has no link that pre-fills a prompt, so the copy button is the handoff.