ByteDance's Seedance 2.5 shipped on 31 July 2026 with native 30-second clips, up to 50 combined references and local in-scene editing, at some of the lowest per-second prices in the category. It is the closest thing to generating a complete short-form ad in one pass. Here is what it does and where it still needs care.
Seedance 2.5 is ByteDance's current flagship video generation model, released 31 July 2026 in the Jimeng and Dreamina apps with a developer API on 7 August 2026. It generates 4 to 30 second clips from up to 50 combined text, image, video and audio references, with synchronized audio, strong character and product consistency, and local in-scene editing that fixes one detail without regenerating the clip. Native output is 480p or 720p, and pricing runs roughly $0.10 to $0.30 per second.
For marketers the 30-second native length is the point: a full TikTok or Reel from one generation instead of stitched 8-second pieces, at a price that makes volume testing realistic.
| Maker | ByteDance (Seed team) |
|---|---|
| Released | 31 July 2026 in Jimeng and Dreamina; API 7 August 2026. Supersedes Seedance 2.0 from February 2026 |
| Max clip length | 4 to 30 seconds (Seedance 2.0 capped around 15) |
| Resolution | Native 480p and 720p; higher resolutions from resellers are upscales |
| Native audio | Yes, synchronized with generation, including beat-aware sync |
| Modes | Text-to-video, image-to-video, multimodal reference-to-video with up to 50 references |
| Consistency | Strong, via role-based asset tagging and reference-guided motion |
| Editing | Local in-scene editing: fix one detail without a full re-render |
| Aspect ratios | 16:9, 9:16, 4:3, 3:4, 21:9 and 1:1 at 24fps |
| Pricing | Roughly $0.10 to $0.30 per second by provider and resolution; Dreamina offers a small free trial |
Versus Seedance 2.0 from February 2026: maximum length grew from about 15 to 30 seconds, references expanded to up to 50 multimodal inputs, and the model gained local in-scene editing instead of requiring full regeneration for small fixes.
Consumer access via ByteDance's Jimeng and Dreamina apps (commercial licensing included on paid Dreamina tiers); developer access via BytePlus and third-party resellers such as PiAPI and WaveSpeed. Dreamina itself has no direct public API.
The native 30-second length plus low per-second cost makes Seedance 2.5 the model for producing a complete short-form ad in one generation, and local editing means a wrong logo colour or an off face is a small fix rather than a full re-render that burns credits.
The risks are real. Seedance went viral in China for uncannily accurate celebrity and copyrighted-character lookalikes, which is a genuine IP and brand-safety exposure. Native resolution tops out at 720p; anything sharper is a third-party upscale. And documentation is scattered across BytePlus, Dreamina and resellers with no single English hub.
Generate anywhere, publish natively to TikTok, Reels and Shorts, attribute signups and sales to every post.
Generate the 30-second hero ad in Seedance 2.5, then upload it to Fastlane to schedule weeks ahead and publish natively to TikTok, Instagram Reels and YouTube Shorts, with signups and sales attributed to each post.
Fastlane fills the rest of the calendar from your website URL with AI UGC videos featuring 1,000+ characters, slideshows, hook + demo videos and memes, remixed from 25,000+ trending videos. Free plan, no credit card. Read our full Seedance 2.5 guide for the workflow in detail.
ByteDance's flagship video generation model, released 31 July 2026, producing up to 30-second clips from up to 50 multimodal references with synchronized audio and local in-scene editing.
Roughly $0.10 to $0.30 per second depending on the provider and resolution. Dreamina offers a small free trial of four generations.
Native 480p or 720p. Higher resolutions offered by resellers are upscales, not model output.
It is available through Dreamina, which is linked to CapCut, with commercial licensing on paid tiers.
2.5 doubled the maximum length to 30 seconds, expanded references to 50 inputs, and added local editing so small fixes no longer need a full regeneration.
Specs verified against the sources above on September 3, 2026. Model access and pricing change often; check the maker's page before you build on any number here.
Join 50,000+ founders and marketers publishing the same video to TikTok, Instagram and YouTube. Free to start.
