ByteDance: Seedance 1.0 Pro

Seedance 1.0 Pro generates a video from text or an image-guided scene. The selector doubao-seedance-1-0-pro-250528 supports both input modes, so the starting point can be a written concept or an existing visual composition. Use a text brief to create a new scene, or provide an image when the subject, arrangement and opening view should be established visually. This 1.0 profile produces silent video; synchronized sound belongs to the separately described 1.5 Pro profile.

For text-to-video, specify the subject, environment, movement, framing and camera in one coherent scene. The documented example moves gradually from a field of daisies to a close view of a flower with dew on its petals. That structure is useful for other reveals: establish the setting, move toward the subject and finish on the detail that matters. For a product concept, describe the material and lighting as well as the action, and give the camera an achievable route through the short shot.

For image-to-video, place the reference in an image_url content item and explain what should change from that starting view. A prepared portrait can define the person and crop while the brief describes a small turn, moving hair or a gentle camera pullback. A product still can establish the object and background while the prompt asks for a controlled reveal. Keep the requested movement compatible with the source composition; an image of one side of an object does not supply an unseen accurate view of every other side.

The image field accepts an accessible URL or a data URL containing Base64 image data. The documented formats are JPEG, PNG, WEBP, BMP, TIFF and GIF. Images must be smaller than 30 MB, with a width-to-height ratio strictly between 0.4 and 2.5 and each dimension strictly between 300 and 6000 pixels. Prepare a clear source at the intended crop rather than relying on a thumbnail whose defining details are already missing. For image-to-video generation, the guide recommends adaptive framing to follow the reference proportions.

The documented 1.0 duration is two to twelve seconds and the frame rate is 24 fps. The output controls include 480p, 720p and 1080p tiers, with 1080p as the Pro default; the guide excludes 1080p in the reference-image scenario. Fixed shapes include 16:9, 4:3, 1:1, 3:4, 9:16 and 21:9. Choose an image-compatible combination for an image-led shot rather than transferring every text-to-video setting unchanged. Inspect the actual returned framing before placing the clip in a layout.

Seed and watermark controls are documented alongside camerafixed. The camera-fixed instruction is not supported in the reference-image scenario, and where it applies the guide does not guarantee a perfectly stationary camera. Prompt-embedded parameter instructions take precedence over separate optional fields. Keep one authoritative setting for duration, resolution and ratio, and make the camera description agree with it. This avoids a scene brief silently overriding the settings selected for the request.

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.

The workflow returns a task identifier first and a completed video location after processing. Preserve the identifier while checking queued or running status; a succeeded task supplies content.video_url and usage information. The documented generated-file retention is seven days. Save the completed clip, source image and exact prompt as the revision record, then inspect subject appearance, movement, image continuity and the closing composition across the whole result.