Sound written with the shot
Dialogue, room tone, footsteps and score are generated in the same pass as the pixels and matched to what is on screen. Set Audio to No audio for a silent render.

Write the shot, or build it from what you attach — 10 images, 5 clips and 5 audio tracks, or one webpage or document.
Dialogue, room tone, footsteps and score are generated in the same pass as the pixels and matched to what is on screen. Set Audio to No audio for a silent render.
Attach up to 10 images, 5 clips and 5 audio tracks on Omni, then @mention each one in the prompt to say what it is for — this face, that room, this track. A mention with nothing behind it holds Generate back.
Paste a link on Omni — a public page, or a PDF, DOC or XLS — and the brief comes from there. It is one link or the attached references, never both.
Frames takes a first frame and fills the motion from there, with an optional last frame to land on. The soundtrack is written as the clip renders.
Attach footage on Edit and describe the change — what to alter, what to leave alone. On this app the clip you bring and the clip that comes back share one 30-second budget, so a longer source leaves a shorter render.
2 to 30 seconds, or Auto to let the model pick a length for the prompt. The Duration select offers 5, 8, 10, 15, 20 and 30 seconds alongside Auto.
Yes. Dialogue, effects and music are produced in the same pass as the video and matched to on-screen action. Set Audio to No audio for a silent render.
Up to 10 images, 5 clips and 5 audio tracks. @mention each one in the prompt — @Image 1, @Video 2 — so the brief says what each attachment is for. A mention with nothing attached blocks Generate until you attach it or take the mention out.
No. It reads one webpage or document, or the attached references, never both — clear one and Generate unblocks. The link needs a full https:// address; a PDF, DOC or XLS path is read as a document, anything else as a web page.
Because on this app the clip you attach and the clip that comes back share one 30-second budget. Twelve seconds of reference video leaves about 18 to render, durations that no longer fit grey out in the picker, and Auto always stays available.
16:9, 4:3, 1:1, 3:4 and 9:16, or Auto to let the model pick — on Omni. Frames and Edit take their shape from the frame or the clip you attach, so no ratio control is shown there.
The same family, returning the same request up to 5x faster, and the panel's default. Same three modes, the same 2 to 30 seconds and the same 480p, 720p and 1080p — at a higher price per second of output than the standard Wan 3.0.