AI video generator
Start from a description, a still image, a pair of start and end frames or a reference video, depending on what the chosen model supports.
Seedance 2.5
Choose a mode, write the prompt, then read the quoted cost before you generate.
Results
Generated content will appear here
How to run a task
Video jobs take longer than image jobs, so the workspace reports progress and keeps the clip in your history once it finishes.
- Sign in and choose a video model from those the catalogue lists as available.
- Pick the mode the model supports and supply the input it needs.
- Set duration, aspect ratio and any sound or render options the model exposes.
- Read the quoted credits, submit, then wait for the job to finish before downloading.
Modes and their inputs
A mode appears only when the selected model declares it, and availability is read from the service catalogue.
| Mode | Input |
|---|---|
| Text to video | a written description |
| Image to video | a start image |
| Start/end frames | a start image and an end image |
| Reference to video | a reference image or video that guides the result |
| Video edit | an existing clip and an instruction |
| Video extend | an existing clip to continue |
Cost and time
Video is charged per job rather than per preview, and long clips or high-resolution modes cost more. The workspace shows the quote for the settings you selected, and the ledger records the charge when the job completes.
Limits worth knowing
A finished clip is a single take, not an editable timeline; changes mean a new job. Frames and reference inputs must be an accepted format, and a job that is still running cannot be downloaded.
Examples
The publisher's first demo clip, thirty seconds at 1280x720. Re-encoded from 40.1 MB to 5.9 MB for this page; frame size and audio are unchanged. Watch the fire: every surface in the shot has to answer to it.
The second demo clip, twenty-six seconds, staged in a bakery. Re-encoded from 26.7 MB to 5.1 MB. The last shot passes through a window with signage, a reflection and a subject on the far pavement in one frame.
Questions
How long does a clip take?
It depends on the model, the duration and how busy the provider is. The workspace shows the job state until it finishes; you can leave the page and find the clip in your history.
Can I make a longer video?
Only with the modes the model offers. Some models include a video-extend mode that continues an existing clip; otherwise a longer result is built from several jobs.
Are sound effects included?
Only when the chosen model exposes a sound option and you enable it. Otherwise the clip is silent.