Seedance 2.5 on Visual Maker AI: Reference-Guided AI Video
How to use Seedance 2.5 for text, image, first/last-frame, and multimodal reference video generation in Visual Maker AI.
Seedance 2.5 is the video model to choose when the important part of the brief is continuity. It can begin from a written description, a still image, first and last frames, or a collection of image, video, and audio references.
On Visual Maker AI, Seedance 2.5 is available in the Video generator. The current interface exposes its supported 480p and 720p output settings, six aspect ratios, and durations from 3 to 30 seconds.
Pick the right starting point
Start with Text to video when the concept is still only an idea. State the subject, action, location, light, and camera movement in that order. A short but directed prompt normally gives the model a clearer shot plan than a list of unrelated style words.
Use Image to video when the opening appearance matters. This is useful for animating an approved campaign visual, product render, or character concept while keeping that image as the visual anchor.
Choose First and last frame when a transition needs a defined beginning and ending. Supply one image for a controlled opening, or two images to specify the final composition as well.
Choose Reference to video when a creative direction depends on several assets. Visual Maker AI accepts separate image, video, and audio references for this mode. Keep each reference purposeful: use images for identity and art direction, video for motion language, and audio for the sound direction. The model supports up to 30 image references and up to 10 video and 10 audio references; submitted reference video and audio must each stay within the supported duration limit.
A dependable prompt structure
For a product video, try a structure like this:
[Product and material] in [place]. [Action]. [Camera movement]. [Light and visual treatment]. [Audio direction, if relevant].
For example: "A matte silver travel camera on a wet basalt pedestal. Water beads roll across the body as the lens extends. Slow macro dolly from left to right. Cool dawn light, restrained premium commercial style, soft rain and distant surf."
That tells the model what must remain recognizable, what must move, and how the scene should be filmed. It is stronger than simply asking for "a beautiful camera ad."
When Seedance 2.5 is a good fit
- A brand needs a short launch clip that continues an existing campaign image.
- A creator needs recurring characters, props, or color treatment across a scene.
- A storyboard needs a guided transition between two compositions.
- A video needs visual references and audio direction in one request.
The result still benefits from iteration. Make one change at a time: first lock the subject and composition, then tune motion, then adjust atmosphere or sound. This makes it much easier to learn which instruction changed the output.
Generate with Seedance 2.5
Open the image-to-video guide and generator to create a video in the same page, or go to Generate and choose Seedance 2.5 from the Video controls.