Wan 3.0 Prompts: Camera Moves, Product Shots, Dialogue & Anime

👁 124 reads

Wan 3.0 prompts work best when you describe shots the way a cinematographer does: framing, camera movement, lighting and sound. On September 26, 2026, we tested four typical use cases: camera movement, product advertising, non-English dialog, and anime-style animation. Each clip cost $0.425 through OpenRouter. Below you’ll find the complete prompts, videos, and what we learned from them.

How to Structure a Good Prompt

After testing this model and the first series, we always structure prompts in the same order:

1. Framing and camera: “close-up”, “medium shot”, “static camera on a tripod”, “slow dolly-in”, “orbit shot”.
2. Subject and action: one main action for 5 seconds. With two or three actions, the model usually shortens or skips the last one.
3. Location and lighting: “courtyard at golden hour”, “dark studio, thin beam of light”.
4. Sound: “natural sound of the kitchen”, and dialog in exact quotation marks with a note specifying the language.
5. Restrictions: “no text, no logo” if you don’t want random text appearing.

We write prompts in English, with only dialog lines in the target language. This is what we did in all tests and the model didn’t mix languages.

1. Camera Movement: Orbiting a Bicycle

Orbit shot: the camera slowly circles 180 degrees around a vintage red bicycle leaning against an old brick wall in a Krakow courtyard at golden hour, ivy on the wall, shallow depth of field, cinematic.

The movement is smooth and convincing, just shorter than requested. Our take: for 5-second clips, it’s better to ask for a “slow quarter orbit”, and save 180-degree orbits for 10-second clips.

2. Product Advertisement

Premium commercial product shot: a matte black wireless speaker on a stone pedestal in a dark studio, a thin beam of light sweeps across it from left to right, subtle reflections, slow push-in, no text, no logo.

The most “commercial” clip in the series. The model didn’t add any brand or text, and the light sweeps exactly as specified in the prompt.

3. Non-English Dialog

Close-up of a smiling grandmother in a cosy kitchen with a pot of soup on the stove. She looks into the camera and says in Polish, clearly: "Najpierw zupa, potem deser." Warm afternoon light, natural sound of the kitchen.

We verified the audio track automatically using Whisper (medium). The result was exactly “Najpierw zupa, potem deser”. An interesting finding from the same test: for clips without speech, Whisper “heard” sentences that weren’t there, such as “Dzięki za oglądanie!“ (“Thanks for watching!”). Automatic captions for scenes without dialog shouldn’t be used without verification.

4. Anime-Style Animation

Hand-drawn 2D anime style: a girl in a white dress rides a bicycle along a country road through golden wheat fields on a summer afternoon, wind in her hair, big white clouds, gentle side-tracking camera.

We described the style with words (“hand-drawn 2D anime style”) rather than a studio name. With image generators, naming a specific studio could result in added watermarks, as we showed in our Z-Image Turbo test.

Frequently Asked Questions

How do you write a prompt for Wan 3.0?

In this order: framing and camera movement, main action, location and lighting, sound, restrictions. Prompt in English, dialog lines in the target language in quotation marks with a note specifying the language.

Does Wan 3.0 handle non-English dialog?

Yes. In our Polish-language test, the model spoke “Najpierw zupa, potem deser” exactly as specified in the prompt, confirmed by Whisper transcription.

How much does one clip cost?

5 seconds at 720p with sound cost us $0.425 through OpenRouter. Details are in our article how much Wan 3.0 costs.

// read moreHow Much Does Wan 3.0 Cost? Pricing, Usage & ComparisonGuidesWan 3.0 vs Wan 2.2: Same Prompts, 8 Clips Side by SideBenchmarksGTA 6 Trailer Style in AI: Wan 3.0 vs Seedance and MiniMaxBenchmarks

Kinetyra