Guru

Seedance 2.5: ByteDance’s Video Model Reaches 30 Seconds

GuruHiTech 594 Views ,  ,  ,  ,  ,  ,  ,  , 

ByteDance released Seedance 2.5 on 31 July 2026. The main change is duration: the model generates up to 30 seconds of video in a single request, with audio produced alongside the picture rather than added later.

That matters more than it sounds. Most competing models stop between five and ten seconds long enough for a social post, not long enough for a scene.

Confirmed features

ByteDance’s Seed research team published the details in its official launch post. Here is what is confirmed.

Thirty seconds in a single generation, with the option to extend the video across multiple further passes while keeping characters and environment consistent.

Up to 50 multimodal references in one request: 30 images, 10 video clips and 10 audio clips. This is the most useful feature for professional work, because it is how you keep a product, a face or a logo stable across a sequence.

Timestamp-level editing. You can change a specific moment in the clip, in both video and audio, without regenerating everything.

Extra editing modes: green screen, camera perspective changes and reference-based editing. The model also supports clay render, motion and creative references. Clay render lets you rough out a scene in basic 3D and use it as spatial guidance for framing.

Audio and video generated together in one pass, an approach meant to remove the sync problems that separate pipelines introduce.

Where you can use it

At launch, Seedance 2.5 arrived on Jimeng AI and the Pro tier of Doubao, two ByteDance apps aimed mainly at the Chinese market. API access was listed as coming to BytePlus ModelArk.

The API is now live. The BytePlus product page lists the Dreamina Seedance 2.5 API as available, with 480P and 720P output and durations from 4 to 30 seconds, sold in three plans (Light, Production and Premium) described in estimated seconds of 480P video.

Outside China, most users reach the model through Dreamina, ByteDance’s CapCut-linked creative platform, or through independent web front-ends. Services like Seedance 2.5 bring text-to-video, image-to-video and reference-guided editing into a single interface with subscription plans. These are not ByteDance-operated products, so check output resolution, watermark policy and commercial licensing before subscribing.

What is not confirmed

Some specifications circulate widely but are not supported by ByteDance’s technical post.

Resolution. The launch post states no maximum resolution or frame rate for Seedance 2.5, the BytePlus API lists 480P and 720P. The Dreamina product page separately promotes 4K, 60 fps and a beta long-video mode reaching 180 seconds marketing claims tied to one consumer surface, not model specifications from the research team, and higher than what the developer API currently offers.

Price. ByteDance has not published a per-second rate for Seedance 2.5. The available references cover the Seedance 2.0 family, where the mini variant was billed at 40 percent of the standard rate and the fast variant at 75 percent. Applying those figures to 2.5 would be an estimate, not a fact.

Independent benchmarks. There was no leaderboard position for Seedance 2.5 at launch. The Decoder reported that Artificial Analysis placed the previous version, Seedance 2.0, at the top of its image-to-video leaderboard among models with audio. That is a signal about the family, not a measurement of this release.

How it compares with Google

The closest comparison is Gemini Omni Flash, announced by Google on 30 June 2026. It generates video from text, images and video with conversational editing, currently reaches clips of about 10 seconds, and costs $0.10 per second of output.

On duration, Seedance 2.5 is clearly ahead. On cost transparency, Google is ahead, because a published per-second rate lets you price a project before you start it.

Conclusion

Seedance 2.5 is most interesting for people who actually finish video projects rather than for people chasing a striking clip. Thirty seconds is enough for a whole scene, fifty references hold continuity together, and timestamp editing means one bad detail does not cost you the take.

The limits are real. The strongest surfaces still target the Chinese market, the developer API starts at lower resolutions than the marketing suggests, and there is no public rate card. Before building a workflow around it, run a paid test and measure your true cost per finished second of video.

Ti potrebbe interessare:
Segui guruhitech su:

Esprimi il tuo parere!

Ti è stato utile questo articolo? Lascia un commento nell’apposita sezione che trovi più in basso e se ti va, iscriviti alla newsletter.

Per qualsiasi domanda, informazione o assistenza nel mondo della tecnologia, puoi inviare una email all’indirizzo [email protected].

Condividi l'articolo

Scopri di più da GuruHiTech

Abbonati per ricevere gli ultimi articoli inviati alla tua e-mail.

0 0 voti
Article Rating
Iscriviti
Notificami
guest
0 Commenti
Più recenti
Vecchi Le più votate