What is actually confirmed?
Fal now publishes callable Seedance 2.5 endpoints for text-to-video, image-to-video and reference-to-video. Saga tested that API surface before enabling it.
The current Saga integration supports 4-30 second output, 480p or 720p resolution, native audio and multimodal references.
Seedance 2.5 became available in Saga Studio AI on 8 August 2026 through a tested Fal API integration.
Release timeline
Seedance 2.0 officially launches
ByteDance announces multimodal text, image, audio and video inputs, plus up to 15-second audio-video output.
Official 2.5 pages appear
CapCut and Dreamina publish the planned 2.5 capabilities that preceded Saga's provider launch.
Seedance 2.5 launches in Saga
Saga validates Fal's key, queue and output path, then opens the model for generation.
Advertised Seedance 2.5 improvements
CapCut advertises a broader product surface. Saga currently exposes the limits published and tested through Fal's API.
- 30 secNative single-generation video, reducing the need to stitch short clips.
- 50 refsImages, video, audio, scripts and other multimodal direction in one workflow.
- 720pHighest resolution currently exposed by Saga's tested Fal API path.
- Local editChange selected regions without rebuilding the entire video.
- ControlStronger character continuity, scene consistency and camera direction.