AI Video Generator
Generate a clip from a written prompt, or supply reference images, a reference clip or a first frame so the subject, the motion and the opening frame follow your own material. Both modes run in the same composer.
- Prompt, references or frames
- Subject consistency from references
- Up to 1080p
- Watermark-free MP4
Describe the shot — subject, setting, light, and how the camera moves. Reference material locks a person, a product or a camera move.
How to generate a video
Choose a mode and add your material
Reference to video takes up to nine reference images and one reference clip, all optional. First and last frame needs an opening still. Everything goes through the same upload slot and is routed by file type.
Describe the motion and the camera
Appearance is already defined by the reference material, so the prompt is best spent on what moves and how the camera behaves. Type @ to point a sentence at a specific image or clip.
Change one parameter, then regenerate
Model, resolution and duration each affect the result independently. Changing one at a time keeps the difference attributable.
Common use cases
Product motion
Use a single packshot as reference to generate a slow push-in or an orbit for a product page or an ad placement.
A consistent character across clips
Supply several angles of the same person so they stay recognisable through the clip and across the ones that follow.
Inserts and B-roll
Cover a missing transition in an edit — a street, a close-up, an overhead shot — without a shoot.
Vertical and multi-format delivery
Generate directly in the aspect ratio you publish in, such as 9:16 or 4:5, instead of cropping afterwards.
Available models
The parameter bar follows the selected model and keeps only the resolutions, durations, aspect ratios and end-frame slot that model actually supports.
Seedance 2.0 Pro & Fast
Pro leads on fidelity and on holding a face together across a shot; Fast gives up the 1080p tier for a quicker turnaround. Both accept an end frame.
- 4–15s
- Pro up to 1080p
- End frame
HappyHorse 1.1
The widest aspect coverage on Labnana, including 4:5 and 5:4, and the only model that will render a three-second clip. It does not take an end frame.
- 3–15s
- 720p / 1080p
- 8 aspect ratios
| Model | Duration | Resolution | Aspect ratios | End frame | Best for |
|---|---|---|---|---|---|
| Seedance 2.0 ProByteDance | 4–15s | 480p / 720p / 1080p | 6 | Yes | Fidelity, people |
| Seedance 2.0 FastByteDance | 4–15s | 480p / 720p | 6 | Yes | Drafts, iteration |
| HappyHorse 1.1Alibaba | 3–15s | 720p / 1080p | 8, incl. 4:5 and 5:4 | No | Short clips, vertical |
Frequently asked questions
Every mode in one composer
Describe the shot, add reference material if you have it, and leave the page while it renders.