GotoRampVideo

Models

The Seedance line, through one API.

Four models share one request format and one balance. Pick by the job: explore cheaply, iterate quickly, produce reliably, finish with the newest model.

Compare

Which model for which job.

Most teams draft on Mini or Fast and render the approved version on 2.0 or 2.5. Switching is one field.

Model IDUse it forTrade-offTypical settings
seedance-2.0-miniStoryboards, prompt tests, bulk draftsLowest cost per second; use it to find the shot, not to deliver it480p · 5 s · audio off
seedance-2.0-fastIterating on framing, motion and timingFaster turnaround than 2.0 at a lower rate480p–720p · 5–8 s
seedance-2.0Ads, social clips, product videoThe workhorse: full multimodal references and native audio720p · 8–15 s · audio on
seedance-2.5Hero shots and final rendersNewest model in the line; highest rate720p · length to fit the edit

Model IDs are GotoRamp's own and stay stable when the underlying model version changes; we announce every change in advance. Per-model limits (resolutions, reference counts, audio) are listed in the API reference.

Capabilities

What the Seedance 2.0 line can do.

Every mode below runs through the same endpoint. You choose it by what you attach to the request.

Text to video

Describe the subject, action, camera and mood. The model returns a clip at the length and format you ask for.

Image to video

Animate a packshot, a keyframe or a still. Add a last frame to control where the shot lands.

Multimodal references

Attach reference images, short video clips and audio to steer identity, motion, camera language and timing.

Native audio

Ambient sound, effects and music generated with the picture, in sync. Turn it off when you score in the edit.

Six aspect ratios

21:9, 16:9, 4:3, 1:1, 3:4 and 9:16, so one prompt covers cinema, feed, story and marketplace placements.

4 to 15 seconds

Set the duration per job. Billing follows the seconds returned, so a 5-second draft costs a third of a 15-second render.

Switching models

Draft on Mini, deliver on 2.5.

Keep the prompt and references; change the model and resolution. Nothing else in your code moves.

Outputs from different models won't be frame-identical. Treat the draft as direction, then approve the final render.

Python
PROMPT = "Rattan lounge chair turning on a white sweep, soft studio shadow"

draft = {"model": "seedance-2.0-mini", "prompt": PROMPT, "duration": 5,
         "metadata": {"resolution": "480p", "generate_audio": False}}

final = {**draft, "model": "seedance-2.5", "duration": 12,
         "metadata": {"resolution": "720p", "ratio": "1:1"}}

Start with a small balance. Scale when it works.

We're onboarding early-access accounts in small batches. Tell us what you're making; we reply within one business day.