One series, three versions
Seedance is ByteDance's video generation series — version 1.0 launched in June 2025, 1.5 Pro in December 2025, and 2.0 on February 10, 2026. The three versions aren't a simple "newer is better" progression; each targets a different use case.
Note: video generation on Nano Banana is currently in demo mode (placeholder results, no credits charged). The prompt structures below are the ones that worked in our testing and carry over directly once real rendering lands.
One-line summary for each version
- Seedance 1.0 Pro: fast, flexible aspect ratios, low cost. For high-volume e-commerce short clips.
- Seedance 1.5 Pro: audio-visual co-generation, beat sync, lip match. For music-synchronized or dialogue content.
- Seedance 2.0: 2K multi-shot narrative, best physics simulation, character consistency across shots. For premium brand films and narrative ads.
Seedance 1.0 Pro — prompt structure
English prompts outperform Chinese, especially for camera movement and lighting descriptions. Duration guide: fast-paced content 5–8 seconds, showcase content 8–12 seconds.
- [subject description] + [action description],
- [background description] + [atmosphere],
- [camera movement].
- [aspect ratio: 16:9 / 9:16 / 1:1], [duration: 5-12]s.
Seedance 1.5 Pro — audio-visual sync
Upload your product image as @Image1 and a music file as @Audio1. The model automatically adjusts visual pacing to match the beat. Rhythmically consistent music (EDM, pop) syncs better than slow ambient tracks.
- [Scene description]. @Image1 [product/person action description].
- Sync visuals to the beat of @Audio1. [Visual style].
Seedance 2.0 — multi-shot narrative
Critical: keep prompts under 60 English words, and 3–4 shots gives the most consistent results. Beyond this length, the model starts ignoring later descriptions. Without explicit shot descriptions, the model decides its own cutting rhythm — which rarely matches commercial editing conventions.
- [Overall scene description]. Generate as multi-shot narrative:
- Shot 1 (wide shot): [wide description],
- Shot 2 (medium shot): [medium description],
- Shot 3 (close-up): [close-up description].
- 15 seconds, 2K resolution.
Seedance 2.0 — single-image 3D reconstruction
Upload a front-facing product image. The model infers the product's 3D form and generates an orbiting camera shot. Useful for products where you need all-angle coverage but don't have a 3D model.
- Single image 3D reconstruction of [product].
- Camera orbiting 270 degrees, revealing all angles,
- [lighting: dramatic side lighting / soft studio light],
- ultra-detailed, 2K, 12 seconds.
Common issues
Unnatural motion? Reduce scene complexity, or specify the motion type explicitly (don't just write "move" — write "slow left-to-right pan").
Seedance 2.0 taking too long? Versions 1.0 and 1.5 are noticeably faster. If time is a constraint, use those for quick iteration to confirm your direction, then use 2.0 for the final output.