Shorts series: a season of 30-second episodes with Seedance 2.5
Shorts series are rolling out. A season of eight 30-second episodes costs $138.72 on Seedance 2.5 at 720p and $30.00 on Wan 3.0. The math and a batch loop.

A season of eight 30-second vertical episodes at 720p costs $138.72 on seedance-2.5 and $30.00 on wan-3.0 on Sume, before retries. Orthotropy's October 2026 platform roundup, read 2026-10-06, says YouTube Shorts series (seasons, episodes, custom thumbnails, sequential playback) have been rolling out since 2026-09-23 on web, mobile and TV. A season is the natural unit to budget for, so this post prices it.
The rollout note is on Orthotropy. Both models take 4 to 30 seconds (Wan from 2) in the Video Router docs.
Season cost, 9:16 at 720p
Seedance's price depends on frame size and length in tokens, so the 9:16 figure uses a 720 by 1280 frame. Wan is a flat rate per second.
| Model | Per episode | Season of 8 | Season of 12 |
|---|---|---|---|
wan-3.0 720p | $3.75 | $30.00 | $45.00 |
wan-3.0 1080p | $7.50 | $60.00 | $90.00 |
seedance-2.5 720p | $17.34 | $138.72 | $208.08 |
seedance-2.5 1080p | $42.65 | $341.20 | $511.80 |
Keep the look steady across episodes
A series lives or dies on continuity. Seedance 2.5 accepts reference images, videos and audio, so you can send the same character image with each episode prompt. It also takes first and last frames, which lets you start episode two on the final frame of episode one. Wan 3.0 accepts references as well. Both are in the catalog's supported_input_references, which you can read from /v1/videos/models.
Generate a season
Submit all eight, each with an idempotency key so a retry never double-bills, then poll as usual.
import os, requests
H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
beats = ["Ep 1: the rooftop garden appears", "Ep 2: the first storm"]
for n, beat in enumerate(beats, 1):
r = requests.post(
"https://api.sume.com/v1/videos",
headers={**H, "Idempotency-Key": f"season1-ep{n}"},
json={"model": "wan-3.0", "prompt": beat + ", vertical, soft light",
"duration": 30, "resolution": "720p", "aspect_ratio": "9:16"},
timeout=60,
)
print(n, r.status_code, r.json().get("id"))Budget the retries
Plan on rerunning one episode in three while you tune the prompt. At that rate an eight-episode Wan season at 720p is about $40.00, and the Seedance equivalent about $184.96. The retry budget covers jobs that finished but that you did not like. Rules for how episodes and seasons behave on the platform are in the Shorts series rules post, and the episode-length batch post covers the format loop.
Cost by the minute of finished series
Eight 30-second episodes is four minutes of video. At Wan 720p that is $30.00 or $7.50 a finished minute. Seedance 2.5 at 720p is $34.68 a minute. The gap is the price of Seedance's wider reference support and first and last frames. If your series depends on a recurring character locking exactly, test both on episode one and decide on what you see, not on the rate.
Budget for the thumbnails and any burned-in captions separately, since the series feature takes custom thumbnails. A still frame from the render is the simplest thumbnail source.
A realistic production plan
Write all eight beats first, one line each, then generate episode one at 480p to check the prompt. At 480p, a 30-second Wan clip is $1.88, so tuning is cheap. Move to 720p for the keeper and repeat for the rest. Keep the character reference image and the style words identical across prompts, and change only the beat. Review the season as a sequence, because episodes that look fine alone can jump in color when played back to back.
Finally, plan the cadence. A series rewards regular releases, so finish the whole season before you publish episode one. That way a failed job never leaves a gap in the schedule.
If the series uses a voiceover, generate the narration separately and send it as an audio reference, so the same voice carries across episodes. Keep the narration under the limits listed for the model, and check the combined audio length before you submit. Episodes that share a voice, a color palette and a recurring character feel like one show, and that is what a series viewer expects.
Sources
Related posts
More in Use cases
- AI album cover generator: square art at 3000×3000
Generate square album art, then upscale: Apple recommends at least 3000×3000. On Sume, generate 2400×2400 and upscale it 1.25× to reach 3000×3000.
- AI avatar for online course videos: build and update lessons
Use an AI avatar as your online course instructor: one reusable avatar, a short talking video per section, captions, and one Timeline join per lesson.
- Talking avatar for PowerPoint presentations, slide by slide
Make a talking avatar presenter for PowerPoint: one Sume clip per slide, up to 60 seconds each, in 16:9 or 4:3 to match the slide, inserted as MP4.
- AI avatar for YouTube videos: Shorts and long-form
Use an AI avatar in YouTube videos: a 9:16 talking video of up to 60 seconds for a Short, or 16:9 segments joined into one long-form video.
Written by Sume