Seedance 2.0

Overview

Seedance 2.0 Fast is a multimodal video generation model in ByteDance's Seedance 2.0 series, built for fast turnaround. It supports three input paths: text-to-video, image-to-video, and multimodal reference-to-video. Developers can generate short videos of 4–15 seconds at 480p or 720p with a single prompt, specify aspect ratio, first and last frames, and reference images/videos/audio, and optionally enable native audio and web search. It is suited to scenarios such as ad previews, social short clips, and narrative segments, where teams need to repeatedly generate and filter usable shots at low waiting cost.

Key Highlights

• Three input paths cover common generation methods — a single model supports text-to-video, first/last-frame image-to-video, and multimodal reference-to-video at the same time, so there is no need to switch between models when choosing an approach.
• Reference materials can be combined to control results — up to 9 reference images, 3 reference videos, and 3 reference audio clips, with a combined duration of no more than 15 seconds, using materials to constrain the visuals and audiovisual style.
• Native audio can be toggled — native audio is generated by default and can be turned off directly when pure visual material is needed for post-production dubbing.
• Text-to-video supports web retrieval — in text-to-video scenarios, web search can be enabled to supplement the prompt with time-sensitive information needed for generation.

Capabilities

Multimodal reference-to-video: constrains generated results through combined input of reference images, reference videos, and reference audio, with up to 9 images, 3 videos, and 3 audio clips per request, suited to shot production that requires material alignment.

First and last frame control: image-to-video accepts a first-frame image and an optional last-frame image to specify the starting and ending frames of a shot, making it easy to generate transitional content between existing materials.

Adjustable duration and resolution: output duration can be set in whole seconds between 4 and 15 seconds, and resolution can be 720p or 480p; 720p offers higher quality, while 480p is faster and more economical. Billing is calculated by output seconds.

Aspect ratio and audio configuration: aspect ratio supports seven options: 1:1, 4:3, 3:4, 16:9, 9:16, 21:9, and adaptive. Native audio can be generated, allowing you to choose the aspect ratio and audiovisual format based on the delivery channel.

Text prompt driven: prompts can be up to 20,000 characters and support Chinese and English. Describe the shot content in text to generate video directly, with no reference materials required.

Use Cases

• Ad teams rapidly generating multiple versions of copy-driven short clips — batch-generate 480p or 720p short clips from text scripts plus reference images for creative pitches and initial screening of delivery assets.
• Storyboard previews requiring consistent characters or style — pass in the same set of reference images and reference videos to generate stylistically coherent short shots as pre-production references for formal shooting or post-production.
• Vertical short video output for social platforms — choose a 9:16 aspect ratio and enable native audio to directly generate content assets suited to vertical feeds.
• Filling in transitional shots between existing materials — specify first-frame and last-frame images to generate the intermediate motion picture, used in editing to bridge two live-action or composited materials.

Limitations and Selection Advice

Single output duration is capped at 15 seconds, making it suitable only for short shots; long videos need to be generated in segments and then stitched together. Reference materials are limited by count and total duration, so complex projects need to be broken down into multiple calls. Resolutions above 480p/720p are not supported, and high-quality delivery requires separate upscaling. Web search only works in text-to-video and is unavailable in image-to-video and reference-to-video scenarios. If higher quality or longer shots are needed, other models in the same series should be evaluated.