ByteDance: Seedance 1.5 Pro

Seedance 1.5 Pro generates video from text or an image and can generate sound together with the picture. The exact selector is doubao-seedance-1-5-pro-251215. Its audio control distinguishes it from the silent Seedance 1.0 profiles: generate_audio=true requests synchronized audio, while false requests a silent video. Use a written scene for a new concept or an image to establish the opening subject and composition.

Describe the visual action and sound as parts of the same scene. Identify the subject, setting, movement and camera, then explain the intended voices, sound effects or background music. A portrait scene might show a person opening their eyes as the camera slowly pulls back, with wind in the environment. For a product scene, connect the sound to the visible event, such as a lid opening or an object touching a surface, instead of adding an unrelated list of effects.

The model can generate voices, effects and background music from the prompt and visual content. The guide recommends putting spoken dialogue inside double quotation marks. Keep speech short enough for the requested action and pauses, and distinguish who speaks when more than one character is involved. Review the actual wording, voice attribution and synchronization with the picture. A sound instruction guides the generated performance; it is not a separate uploaded voice-reference input in this profile's documented request.

Image-to-video uses an image_url item alongside the text. Supply an accessible image URL or a Base64 data URL in JPEG, PNG, WEBP, BMP, TIFF or GIF format. The documented image limits are below 30 MB, both dimensions strictly between 300 and 6000 pixels and a width-to-height ratio strictly between 0.4 and 2.5. Prepare the intended crop and subject clearly, then describe the movement from that composition. The image example uses adaptive ratio to follow the reference.

The profile-specific duration guide lists whole seconds from four to twelve, or -1 for the model to choose a duration within that range. The default duration is five seconds. Seedance 1.5 Pro defaults to 720p and the retained profile table documents 480p and 720p; the dedicated operation guide does not offer 1080p for 1.5 Pro. Fixed ratio choices include 16:9, 4:3, 1:1, 3:4, 9:16 and 21:9, with adaptive as the documented 1.5 default. Inspect the actual clip's duration and proportions rather than treating the requested values as an already returned artifact.

The request exposes two processing tiers. service_tier=default selects online processing, while flex selects offline processing for work that tolerates more delay. The tier cannot be changed after the task is submitted. Choose it for the workflow's timing needs, and retain that selection in the request record. Audio mode and processing tier are separate choices: a silent clip is not automatically an offline job, and enabling audio does not select a processing tier.

Keep the seed, watermark and camera instructions consistent with the scene. Parameter commands embedded in the text take precedence over separate optional parameters. The fixed-camera control is excluded in the reference-image scenario and is not a guaranteed visual lock where it applies. For an image-led animation, explain the intended movement directly rather than assuming a camera control removes the need to describe the shot.

Usage is measured in video tokens: output width × output height × frame rate × output duration / 1024. The completed task reports this quantity as usage.completion_tokens. Retain the reported usage with the finished clip and its requested settings. Audio enabled or disabled and online or flex processing remain distinct categories in the usage-based charge.

Submission returns an asynchronous task identifier. Follow the task until completion, then save the video supplied in content.video_url. The documented generated-file retention is seven days. The expiry control is measured from creation, defaults to 172800 seconds and accepts values from 3600 to 259200 seconds; an unfinished task becomes expired after its configured interval. Keep the final clip, original image and exact visual-and-audio brief together, then review the full result for action, identity, camera continuity and sound alignment.