Dreamina Long Video Mode: 3 minutes vs Sume's 30 s Seedance

Dreamina's Seedance 2.5 launch release says its standard mode generates up to 30 seconds of continuous video and a Long Video Mode supports videos of up to three minutes. Sume does not list a long-video mode: seedance-2.5 accepts 4 to 30 seconds per request, so a three-minute film on Sume is several requests joined together.

That is a real difference in product shape, not a spec dispute. The sections below say what each side documents, what you can build on Sume today, and what it costs.

What does Dreamina document?

The Dreamina launch release says Seedance 2.5 launched worldwide on July 31, 2026, with global availability on Dreamina, up to 30 seconds in standard mode, up to three minutes in Long Video Mode, and up to 50 reference assets made of 30 images, 10 video clips and 10 audio files. ByteDance's own launch post describes up to 30 seconds per generation and multi-round extensions for longer videos.

Neither page explains how Long Video Mode works internally or whether it is exposed through an API. It is described as a Dreamina feature, so this post does not assume it exists in the model's API.

What does Sume document?

Sume's Video Router docs list seedance-2.5 at 480p, 720p and 1080p for 4 to 30 seconds, and every other catalog model at 15 seconds or less, apart from Wan 3.0 at 2 to 30 seconds and the 5-30 second source-length models Genjutsu and Recast. There is no longer-than-30-second option in the catalog.

How do you make a three-minute film on Sume?

Plan it as six 30-second shots, or a dozen shorter ones, each its own request. Give every shot the same reference images, where the model accepts them, so the character and product stay consistent, and keep the aspect ratio and resolution fixed across shots. Then assemble the clips with Timeline 1.0, which takes one audio spine and up to 200 ordered video slots and returns one MP4.

Two Timeline facts to plan around. It needs every URL to be a media.sume.com artifact or asset of your workspace, so completed Sume clips work and outside files need importing first. And its document has an audio spine (it can be silence) rather than per-clip audio mixing, so check how you want the Seedance audio handled before you rely on it. This post has not tested how clip audio carries through.

What does it cost, and where does it break?

At 720p, a 30-second seedance-2.5 clip comes to roughly $17 by our reading of the per-token rate, so six shots are about $104 (check the real estimate before you submit), plus Timeline at $0.10 per output minute rounded up, which is $0.30 for three minutes. The seams are the cost you do not see: continuity across clips is your job, not the model's. The extension post prices the rounds in detail.

Long Video Mode may keep a scene coherent across a longer span than six independent requests will. If your film needs one unbroken take, a hosted Dreamina feature is the route that documents it. If you need an API, a billing record per shot, or other models in the same film, Sume's route is the joined clips.

Which should you choose?

Choose Dreamina's Long Video Mode if you want a single three-minute generation inside the Dreamina app and can work with its limits. Choose Sume if you need an HTTP API, per-request pricing you can estimate before submitting, and the ability to mix seedance-2.5 shots with other catalog models. For the 2.0 to 2.5 differences, see Seedance 2.5 vs 2.0, and for all clip limits by model, AI video length limits.