Skip to main content

​
What It Is

Video Generation is a built-in superun capability. You can ask superun to create a short video directly in chat, or enable the plugin to build a complete video-generation workflow into your own app. Direct generation currently uses Seedance 2.5. It supports text, image, video, and audio references, synchronized audio, virtual presenters, and authorized real-person videos. App integrations can also use other video models currently enabled for the project. The plugin is enabled per project.

​
Where to Find It

  1. Open any project.
  2. Switch to the Build tab.
  3. Open Skill Library from the left navigation.
  4. Find Video Generation.

​
Current Capabilities

  • Text to video: Generate a video from a written scene or script.
  • Multimodal references: Seedance 2.5 can combine image, video, and audio references, including audio-only reference generation.
  • Native audio: Generate synchronized speech, sound effects, and background music, including virtual talking-head videos.
  • Authorized real-person video: Create real-person appearances or talking-head videos after confirming lawful rights to the person’s likeness and voice.
  • Flexible duration: Seedance 2.5 supports 2–30 seconds or automatic duration. Other models expose their own allowed values in the product.
  • Background generation: Continue working while the task runs.
  • Progress and results: Follow review and generation status from the video card.

​
Seedance 2.5 reference limits

Reference assets guide the model’s understanding of subjects, actions, and style. They are not exact first-frame or last-frame controls.

​
Real people and virtual characters

  • Virtual or AI-generated character: Use the normal reference flow.
  • Real-person photo: Confirm that you have lawful rights to use the person’s likeness and voice, and provide exactly one HTTPS portrait.
Whether the person appears and whether the video includes speech are separate choices. A silent real-person video still requires rights confirmation. For a real-person talking head, enable synchronized audio and include the script in the prompt. superun handles the one-time private asset import on the server and never exposes the provider asset ID. The first submit may therefore take longer than a normal request.

​
Models available to app integrations

Depending on the current product configuration, an app can use models such as:
  • Seedance 2.0 Fast, Seedance 2.0, Seedance 2.0 Mini, and Seedance 2.5
  • MiniMax H3
  • Kling 3.0 and Kling 3.0 Pro
  • Wan 2.7
Each model has different duration, resolution, and reference-input limits. Update the available controls whenever users switch models. The product’s current model selector is the final source of truth.

​
Common Failures

The video card reports failures such as insufficient credits, rate limits, missing access, unsupported duration or reference counts, content review rejection, or missing real-person rights confirmation. Adjust the input shown in the error and submit again; do not create another paid task only to check progress.

​
Disable

Open the plugin detail page, click Disable, and confirm. Existing stored output follows the project’s storage rules.

​
FAQ

Up to 30 images, 10 videos, and 10 audio files, with no more than 50 reference assets in total.
Seedance 2.5 supports 2–30 seconds or automatic duration. Other app models use the values shown after selecting that model.
Yes. Seedance 2.5 can generate synchronized speech, effects, and music. A real-person talking head requires one HTTPS portrait, rights confirmation, and the spoken script in the prompt.
Yes, with Seedance 2.5. You must have lawful rights to the person’s likeness and voice and provide exactly one HTTPS portrait. Virtual or AI-generated characters do not require real-person rights confirmation.