Seedance 2.5 is now available via Partner Nodes
30-second single-shot generation, up to 50 reference assets, prompt-driven video editing, and native output in 10+ languages — Seedance 2.5 lands in ComfyUI as a Partner Node.
Seedance 2.5 is now available in ComfyUI through Partner Nodes. This state of the art model moves the format past the short-clip ceiling most video models still operate under: one run produces a single continuous 30-second segment, driven by up to 50 reference assets.
The other half of the release is control. Seedance 2.5 handles instruction following more reliably than its predecessor, supports editing and extending existing footage, and generates spoken content natively in more than ten languages.
Model Highlights
A full 30 seconds in one pass. Seedance 2.5 generates a single continuous 30-second segment directly — no stitching, no chaining shorter clips and hoping the seams hold. Composition, lighting, and subject continuity carry across the whole duration.
Up to 50 reference assets per run. A single generation accepts 30 images, 10 videos, and 10 audio files. That’s enough to lock character, environment, motion reference, and soundtrack in one prompt rather than negotiating them across multiple passes. References are addressed inline in the prompt, so you can point a specific instruction at a specific asset.
Second-level shot control. Prompts can be written as a timeline. Specify what happens in each time block, call out shot size, camera type, and transitions, and the model holds to the schedule — including hard cuts on the beat. Long-form output is only useful if you can direct it, and this is the mechanism.
Controllable editing and extension. Feed in an existing video and describe the change in plain language. Remove everyone in frame except the protagonist, and the model isolates the subject, reconstructs the background behind the removed elements, and keeps the rest of the shot intact. The same path covers extending footage past its original end point.
Multilingual performance with lip sync. Seedance 2.5 generates dialogue and vocal performance natively in more than ten languages. Lip shapes and jaw movement match the words being spoken, which holds up even when a single clip cycles through several languages in sequence.
Examples
Getting Started
Update ComfyUI to the latest version, or open Comfy Cloud.
Load a template from the Templates panel, or download the workflows below.
Attach your references, write your prompt, and run.
As always, enjoy creating.