How do I make a 180-second Shorts episode from 30-second Wan clips?

Generate six 30-second Wan 3.0 clips, then join them in one Timeline 1.0 render. At 720p that is 180 output seconds at $0.125, which is $22.50, plus $0.30 for a three-minute render.

YouTube's Help page on Shorts series describes a series as a Shorts-only playlist with seasons and episodes, with each episode at most 180 seconds (read 2026-10-07). So 180 seconds is the ceiling for one episode, and Wan's 30-second call limit means a full-length episode is a six-part assembly.

Why six clips

The Videos docs list wan-3.0 at 2 to 30 seconds per call. A scene that must run three minutes is therefore a plan: six beats of 30 seconds, each written as its own prompt, each ending where the next begins. Seedance 2.5 accepts 4 to 30 seconds, so the same arithmetic applies there; use the model catalog for the price of your pick.

The clips are separate jobs, which is the useful part. A bad beat is a re-roll of 30 seconds, not of 180.

Assembling the six beats

Timeline 1.0 takes a sequence of Sume-hosted video slots, transitions and an audio spine. Adjacent video[] slots play one after another, so six clips in order is the whole edit. Render the audio once as a single track instead of trusting six separate soundtracks to line up; the audio spine post covers that.

Captions are priced for clips of at most 60 seconds, so burn them on three 60-second parts or on the finished beats, not on a 180-second master. Check the Video captions page for the current rule.

• Write each beat so its last frame is a usable first frame for the next.
• Send resolution: "720p" explicitly; Wan's estimator defaults to 1080p when it is missing.
• Keep the six job ids in your own table so a re-roll replaces one slot.

When to stop at 60 seconds

Most series do not need three minutes. The cost of 60 seconds is one third of the table above, and the season cost post shows ten of them.

What can go wrong at the seams

The risk of a six-part episode is the joins. Each clip is its own generation, so lighting, face and framing can jump between beats. Two habits help. First, end each prompt on a still moment, such as a held look or a door closing, so the cut hides the jump. Second, pass the last good frame of one beat as the first frame of the next with frame_images, which the Videos docs describe as image-to-video control.

Audio is the other seam. Models with native audio generate a track per clip, and six tracks do not line up on their own. Either mute the generated audio and lay one spine under the whole episode, or detach each track and join them as in the audio post. A single spine also lets the music run across the cuts.

Last, decide the aspect ratio once. Send aspect_ratio: "9:16" on every beat, or the clips will not match in the render. Check the supported ratios for your model in the catalog before the first request.