Seedance 1.5 Pro is ByteDance's video generation model, and doubao-seedance-1-5-pro-251215 is its specific invocation version. It supports text-to-video, image-to-video, and sound generation, making it suitable for turning shot descriptions or still images into short films. When you need to start from a specified image, control the final composition, or add sound to a scene, this version provides a clear creative workflow.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
The values above are the invocation ranges for this version on this platform. Creation modes and controls are used according to this model and do not represent the capabilities of other Seedance versions.
1.5 Pro can enable sound while generating video, so you do not need to split every creation into separate visual and audio tasks. Prompts can describe character actions, environments, and the sounds you want to hear at the same time, such as wind blowing through hair and the sound of wind. When silent footage is needed, turn off sound generation and arrange music or editing afterward.
Image-to-video can begin from a first-frame image, or include a last-frame image to specify the ending shot, then use text to describe the action and camera movement in between. It is suitable for shot designs with an established composition; unlike providing only character references, first and last frames emphasize the beginning and end of the image rather than identity references across scenes.
Landscape, portrait, square, and ultra-wide formats can serve presentations, mobile short videos, and conceptual shots respectively. During invocation, you can explicitly set the aspect ratio and duration, and use a fixed camera or random seed depending on the task. When you need to continue planning the next segment, you can request the last frame to be returned as an image for preparing the following shot.
Use a product image as the first frame, add subject motion, background changes, and camera direction, and generate short shots for advertising edits. For example, make steam rise from a cup and have the camera slowly move closer, rather than merely asking for it to “look premium.” The deliverable is downloadable video footage, making it easy to add brand text and editing rhythm later.
Enter a clear short scene, specifying the subject, action, and sound—for example, rain falling on a street while a person walks by holding an umbrella—and enable generate_audio. This is suitable for first validating creative directions that combine sound and visuals before deciding on subsequent production; after delivery, the audio track should be reviewed, not just individual frames.
When you already have opening and ending storyboard frames, set the two images as the first and last frames respectively, describe the movement, turn, or camera changes in between, and generate a transition preview. It can help directors, animation teams, or design teams discuss whether the shot works; the delivered video can also be used for presentations rather than replacing full-length production.
If the task is a text- or first/last-frame-driven short video and you want to generate sound directly, 1.5 Pro is the clear choice. Compared with the 1.0 series, which does not support generate_audio, it adds a path for sound-enabled creation. If you only need silent visuals, there is no need to enable audio just for the version number; consider the relevant 1.0 models separately for text-to-video and image-to-video tasks.
If you need character reference images, reference audio, or reference video, consider the 2.0 series rather than adding these assets as roles in 1.5 Pro. If you need to edit existing video, extend clips, or generate up to 30 seconds, choose 2.5; if you need 4k output, consider 2.0 Standard. Your choice should follow the task, not just the version number.
Provide text in content; images can use first_frame/last_frame to represent the starting and ending visuals; the last frame should be consistent with the creation method and asset composition.
Specify model=doubao-seedance-1-5-pro-251215 for /seedance/videos; first test with duration=5, resolution=720p, and a clearly defined aspect ratio. When sound is needed, explicitly set generate_audio=true; it is disabled by default.
For asynchronous processing, first obtain the task_id, then query /seedance/tasks or receive a callback; after completion, check the subject, action, ending, and audio, then save the final video and returned last frame as needed.
Use an image by a window on a rainy day as the first frame. The camera slowly moves closer to the glass, raindrops slide down, the indoor lighting is soft, and the sound is rain outside the window, with no voices or music.
Request sound with generate_audio=true, and check the rain sound and raindrop motion segment by segment; do not use the 2.x multimodal reference_image method.
Set generate_audio=true in the request, and describe the scene and the sounds you want to appear in the text. This option is disabled by default, so merely writing a sound description without enabling the parameter should not be considered a sound-enabled delivery solution. After generation is complete, check both the visuals and the actual audio track.
Add an image_url item to the content array, place the address in the image_url.url object, and set the role to first_frame or last_frame respectively. First- and last-frame tasks should also include a textual description of the changes in between; image_url cannot be written directly as a string.
This version is used with 480p, 720p, 1080p, and durations of 4–12 seconds, and also supports automatic duration. For 4k tasks, consider 2.0 Standard; for tasks up to 30 seconds, consider 2.5. Do not directly use the resolution or duration settings of other versions with 1.5 Pro.
You can use a person's photo as the first frame, so the shot begins from that image; however, this differs from a reference_image character reference. If the goal is to preserve the person's identity while changing the scene, consider the 2.0 series that supports this reference method. The 1.5 Pro image workflow focuses on first and last frames.
Submit model and content to /seedance/videos, with the model set to doubao-seedance-1-5-pro-251215. Asynchronous mode returns task_id; then query the task, or use callback_url to receive notifications. After the task is complete, download the video through data.video_url.
Choose a development approach or application suited to this model and follow its setup guide.
↗Python SDK Integration Guide
SDK installation, authentication and examples↗TypeScript SDK Integration Guide
SDK installation, authentication and examples
↗
↗
↗This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs