Which Seedance for a UGC ad: 2.5, 2.0, Fast or Mini on Sume?

Test UGC hooks on Seedance 2.0 Mini or Fast at 480p, then spend on 2.5 only when the ad needs 16-30 s or 1080p. Prices for 15 s vertical clips.

5 min readSume
All posts

For a UGC-style ad, test hooks on seedance-2-mini or seedance-2-fast at 480p, then render the keeper on seedance-2.5 if it needs more than 15 seconds or 1080p. The 2.0 family stops at 15 seconds on Sume; Seedance 2.5 accepts 4-30 seconds at 480p, 720p and 1080p.

A 15-second 9:16 clip at 720p costs $2.84 on Mini, $4.54 on Fast, $5.67 on Seedance 2.0 and $8.67 on Seedance 2.5. UGC testing is about volume, so the cheapest row that still shows the hook is usually the right first pass.

What UGC needs from a video model

A UGC ad is a person-shaped moment: a face close to the lens, a hand holding a product, a short line of speech or room sound, and a vertical frame. None of that needs a long clip. Most hooks are decided in the first few seconds, so the first question is how many variants you can afford, not how long each one runs.

All four Seedance ids in the Sume catalog list text-to-video, first and last frame image-to-video, and image, video and audio references, with audio on. The differences that matter for a UGC workflow are the length ceiling, the resolution ceiling and the price per second. The Video Router docs list the envelopes: 2.5 is 4-30 s at 480p/720p/1080p, 2.0 is 4-15 s at 480p/720p/1080p, and Fast and Mini are 4-15 s at 480p/720p.

Price of a 15-second vertical clip by tier

The table prices the same 15-second 9:16 request on every tier. A dash means the id does not list that resolution in the Video Router catalog.

15 s 9:16 clip, Sume price for the clip, list x 1.25 rounded up to the cent, read 2026-10-05
Model id480p720p1080pMax length
seedance-2-mini$1.32$2.84-15 s
seedance-2-fast$2.11$4.54-15 s
seedance-2$2.64$5.67$12.7615 s
seedance-2.5$4.04$8.67$21.3330 s

A three-step pick

  • Hook test: send 10 to 20 variants to Mini at 480p. A 15-second variant is $1.32 each, so 20 variants are $26.40.
  • Keeper pass: re-render the two or three winning prompts on Seedance 2.0 or 2.5 at 720p. This is where face detail and motion get judged, so do not judge them on the Mini draft.
  • Long cut or 1080p: only Seedance 2.5 reaches 30 seconds, and only 2.5 and 2.0 reach 1080p. Pick 2.5 when the script needs the longer take, 2.0 when a 15-second 1080p clip is enough and you want the lower price ($12.76 against $21.33).

Why a draft is not a preview of the keeper

The tiers are different models, so a prompt that works on Mini is a hypothesis for 2.5, not a guarantee. Sume's catalog marks no seed support for these models, so you cannot lock a take and upgrade it; a re-render is a new take. Budget for that by treating the draft pass as a prompt screen and the keeper pass as a separate purchase.

Seedance 2.5 was announced by ByteDance Seed on 2026-07-31 with 30-second generation and up to 30 image, 10 video and 10 audio references per pass. Sume's own request validation is narrower: it rejects more than 9 reference images, 3 reference videos, 3 reference audio files or 12 references in total. For UGC that is plenty: one product photo, one creator-style frame and a voice clip fit with room to spare.

Send the keeper request

This is a vertical 15-second request on 2.5. Swap the model id to test on another tier; the body is the same.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ugc-keeper-001" \
  -d '{"model":"seedance-2.5","prompt":"Handheld selfie video of a woman showing a moisturizer jar to camera, natural window light, she says one short sentence, ends on the jar label","duration":15,"resolution":"720p","aspect_ratio":"9:16"}'

Where the answer changes

If your ad needs a face to stay consistent across several clips, spend more on reference images than on resolution. If it needs spoken lines, read the model's audio fields in the catalog before you commit, because a UGC clip with a line of speech is judged on sound as much as on picture. For polling and downloads, the video generation docs show the same async flow for every id: submit, poll the polling URL, download from the content URL.

Sources

Related posts

More in Models

All Models posts

Written by Sume