Wan prompt workflow · Wan 2.7
Write Wan multi-shot prompts with clear beats, continuity anchors, camera transitions, sound cues, and an editable ending. Includes reusable Wan 2.7 and Wan 2.6 templates.
A multi-shot prompt works when each beat changes the viewer's understanding. Define what the shot reveals, what stays continuous, and why the next cut or transition exists.
Use explicit labels for shot size, camera path, and transition. Do not accidentally redesign the character or product each time the camera changes.
Visual examples
Copyable prompt templates
Keep the subject, action, camera, light, sound, and ending coherent. Bracketed fields are the variables to replace.
Beat 1: macro detail of [product material], slow lateral slide. Beat 2: pull back as [physical reveal] exposes the full product. Beat 3: controlled orbit and settle on the approved hero angle. Preserve exact geometry, color, and label placement; restrained studio ambience.
Best with a clean product reference image.
Use in generatorSame character and wardrobe throughout. Shot 1: wide tracking shot as [character] enters [place]. Shot 2: medium push-in when they discover [object]. Shot 3: close reaction, subtle room tone, hold the final expression for one second. Consistent time of day and light direction.
Keep the visible discovery simple enough for the duration.
Use in generatorShot 1 opens on [ordinary state]. A continuous [material/weather] transformation begins around the same subject. Shot 2 tracks through the transformation without changing identity. Shot 3 resolves on [campaign state], centered composition, clean negative space, no generated text.
Use first and last frames when the endpoint must be exact.
Use in generatorOne traveler, same face, hair, jacket, and backpack. Beat 1: [location A], wide establishing move. Beat 2: match cut on the traveler's walking direction into [location B]. Beat 3: sunrise close-up at [location C]. Natural environmental sound changes at each transition; stable anatomy.
Repeat the identity block before each beat if drift appears.
Use in generatorBeat 1: [subject] notices [threat], locked medium shot. Beat 2: one readable evasive action, handheld tracking with realistic weight. Beat 3: subject reaches [safe endpoint], camera stabilizes, sound falls to room tone. Same environment and clothing, no extra characters.
One action per beat reduces structural failures.
Use in generatorShot 1: exterior atmosphere of [venue], slow push. Shot 2: hands preparing [relevant object], shallow depth of field. Shot 3: medium portrait of [speaker] looking just off camera, natural room tone, hold for the interview cut. Consistent color grade.
Add dialogue in a separate test when wording must be exact.
Use in generatorGenerated identity, hands, product geometry, text, logos, dialogue, sound timing, and rights require a final human check.
For 5–15 second output, begin with two or three readable beats. More shots leave less time for each action and increase continuity risk.
Use one reference image when available and repeat a short identity block: face, hair, wardrobe, proportions, and must-keep accessories.
Yes. Place dialogue, ambience, or an effect next to the visible event it supports, then review timing and wording in the actual result.