Wan 3.0 prompt generator

The Wan 3.0 prompt formula

[Overall description] + [references: Image N / Video N / Audio N] + [Shot N (start–end seconds): subject + scene + motion + aesthetic control] + [dialogue: X says: “…”] + [sound effects / background music] + [style / mood] + [negative list]. Skip any block you don’t need. The sample scene comes out like this:

A young woman in a red raincoat stops and looks up as the rain starts in a neon-lit Tokyo side street at night. Push in, soft neon lighting.
Shot 1 [0-3s] Wide shot, she walks toward the camera under the neon signs.
Shot 2 [3-6s] Close-up, raindrops land on her face and she smiles.
The woman says: "It finally came."
Sound effects: rain on the pavement, distant traffic.
No background music.
Cinematic 35mm film style, quiet, nostalgic mood.
Avoid: subtitles.

Shots and single takes

Number each shot and give it a time range, for example “Shot 1 [0-3s] … Shot 2 [3-6s] …”. Shots must join end to end, with no gaps or overlaps. Pick one format and use it throughout. Alibaba’s prompt guide suggests 2–5 seconds per shot, while its usage guide says 4–6; the Chinese versions of both pages were last updated on September 24, 2026.

For one unbroken shot, put “Generate single shot.” on the first line. Tick “One continuous take” in the generator and it does this for you.

Say “No dialogue.” or the model decides

Wan 3.0 generates dialogue, sound effects and music natively. If you don’t want speech, write “No dialogue.”: Alibaba’s guide says that otherwise the model decides for itself whether to add lines. Likewise, write “No background music.” if you want none. Write dialogue as character + speech verb + colon + quoted line. Add “Lip sync.” for lip sync, and “Voice timbre reference Audio 1” once to clone a voice from an uploaded clip.

A negative list instead of negative_prompt

Wan 3.0 has no negative_prompt parameter. End the prompt with a short list of what you don’t want. Alibaba’s guide says to list only real exclusions: don’t pad the list and don’t repeat what the positive part already says. Wan 2.7 and 2.6 still have a negative_prompt field of up to 500 characters.

References and prompt rewriting

• Cite uploads as Image 1, Video 1, Audio 1: capital letter, a space before the number, counted separately per type in upload order. With a single reference you can just say “the reference image”.
• Wan 3.0 accepts up to 10 reference images, 5 reference videos (15 seconds in total) and 5 audio clips.
• prompt_extend is on by default: an LLM rewrites your prompt before generation. Alibaba says it helps short prompts and adds latency. It must stay on when the input is a file or a link.

Wan 3.0 at a glance

Related tools

• Wan 3.0 price per secondWan 3.0 price per second: $0.05 at 480P, $0.10 at 720P and $0.20 at 1080P on Alibaba Cloud, audio included. Compare Prime and resellers from fal to OpenRouter.
• Seedance prompt generatorFree Seedance prompt generator. Builds Seedance 2.5 prompts with timestamps and ByteDance’s sound tags, or Seedance 2.0 prompts in Shot 1 / Shot 2 format.
• Kling prompt generatorFree Kling prompt generator. Writes Kling 3.0 prompts in Kling’s own order and splits them into up to six shots with the official “shot n, m, words;” syntax.
• MiniMax H3 prompt generatorWrite MiniMax H3 prompts the way the official cookbook lays them out: one core sentence, a shot-by-shot timeline and exclusions at the end. Works for Hailuo AI.