Wan prompt workflow · Wan 2.7
Write Wan video prompts that connect dialogue, ambience, music, and effects to visible actions without overloading a short clip.
Audio direction becomes clearer when the prompt says who speaks, what makes the sound, and when it happens relative to the action.
Write 'the glass touches the table, then one clean click' instead of only asking for dramatic sound. Temporal relationships are easier to evaluate after generation.
Visual examples
Copyable prompt templates
Keep the subject, action, camera, light, sound, and ending coherent. Bracketed fields are the variables to replace.
Medium close-up of [character] in [room]. They say exactly: '[short line]'. Subtle facial performance, natural room tone, no music, slow push-in, hold after the final word.
Shorter lines leave more time for readable performance.
Use in generatorMacro view of [product detail]. Silence except soft studio room tone. As [part] locks into place, one precise metallic click occurs; a warm rim light activates at the same instant. Pull back to the full product.
One synchronized event is easier to verify than a full soundtrack.
Use in generatorWide tracking shot along a rain-soaked street at night. Layered rain, distant traffic, and one passing bicycle bell. No dialogue or music. Camera follows [subject] while footsteps remain synchronized with visible contact.
Keep ambience sources spatially plausible.
Use in generator[Subject] performs [single action]. A restrained rhythmic pulse builds for three counts. On the fourth count, [visible impact] happens with one deep hit; particles expand, then the soundtrack drops to ambience as the camera settles.
Specify a visible payoff for the audio peak.
Use in generatorNatural interview portrait of [speaker]. Quiet [location] room tone, faint ventilation, no music. The speaker breathes, then says: '[line]'. Stable medium shot, realistic mouth timing, hold at the end.
Review exact wording and lip timing before use.
Use in generatorSlow orbit around [machine] in a dark workshop. Low electrical hum follows the visible indicator lights. As the mechanism starts, layered gears accelerate naturally; no music, no voice, finish with the machine at steady speed.
Match changes in sound intensity to visible mechanics.
Use in generatorGenerated identity, hands, product geometry, text, logos, dialogue, sound timing, and rights require a final human check.
No. Connected model modes can report different audio capabilities. Confirm the selected route in the live generator before relying on a sound workflow.
Use one short line that leaves time for breathing, visible delivery, and an ending hold. Test wording and timing rather than assuming every word will fit.
No. A technical ability to upload or describe a voice does not grant rights to imitate it. Obtain appropriate permission and review applicable terms.