When to use
Almost never. R2V is the universal default for everything UGC. Use i2v ONLY when:
- The user has named
seedance-2.0-i2vexplicitly ("use i2v") seedance-2.0-r2vis unavailable / inactive in the registry (checkquickdesign video models)
There is no scenario where i2v is "simpler" enough to justify silently picking it over R2V. R2V accepts a single --reference-image cleanly — it works exactly like i2v's --image from the user's perspective, and it preserves all the primitives the skill depends on:
@Image1reference grammar- Repeatable
--reference-imagefor multi-product --reference-audiovoice continuity for multi-segment
i2v silently forfeits all three.
Capabilities (for reference)
quickdesign cost seedance-2.0-i2v -d 12 -r 1080p
quickdesign video models | jq '.data[] | select(.slug=="seedance-2.0-i2v")'- Single
--imageinput only — noimage_urls[]array, noaudio_urls[], novideo_urls[] - Same duration grid (4–15s), same resolutions, same aspect ratios
- Native audio: yes, but no voice continuity primitive
How to apply
Don't pick this model autonomously. If considering it, either:
- Default to
seedance-2.0-r2vinstead, OR - Route through
AskUserQuestionand let the user explicitly pick i2v.
See ../SKILL.md cardinal rule #0.