KreadoAI now supports Wan 3.0, Alibaba's video model for a full clip. Wan 3.0 what's new is a 2–30 second range at 30 fps, native dialogue, music, and effects, mixed image, video, and audio references, instruction editing, extension, and a separate first-and-last-frame mode.
Select Wan 3.0 Standard or Wan 3.0 Prime in the video generator. The page is the Wan 3.0 AI Video Generator. The worked prompts are in learn how to use Wan 3.0.
Wan 3.0 vs Wan 2.7
Both models are on KreadoAI, each on its own page. This table uses what each page already publishes. If the Wan 2.7 cell does not list a point, that page does not publish it. It does not mean the older model is incapable of it.
| Wan 3.0 | Wan 2.7 | |
|---|---|---|
| Video length | 2–30 seconds, 30 fps | This 30-second range is not on the Wan 2.7 page |
| Native dialogue, music, and effects | In the same generate, when audio generation is on | Not on the Wan 2.7 page |
| References | Images, videos, and audio together, up to 20 files | Image-to-video, a 9-grid reference, and first and last frames |
| Instruction editing | Change one element and keep the rest | Instruction editing |
| Extend forward, backward, or both | Original plus new seconds stay within 30 | Those three directions are not on the Wan 2.7 page |
| First and last frames | A separate mode from reference uploads | Listed with the other shot controls |
Open Wan 2.7 when the job is the 9-grid workflow on that page. Open Wan 3.0 when the job is a timed clip with sound, a mixed reference set, or a revision of footage you already have.
What can you do with Wan 3.0?
On KreadoAI the same model covers the jobs that used to be split across a generate, an editor, and a sound pass.
Longer clips, written as shots
The ceiling is 30 seconds. A wide shot, a detail, and a closing move fit when the prompt gives each beat a clock range. One vague line spends the length on a single drift.
Native dialogue, music, and effects
Ask for one spoken sentence, the music bed, and the effect, and turn audio generation on. Audio off returns a silent file even when the prompt mentions a voice.
Mixed references
A reference job can combine images, videos, and audio, up to 20 files: as many as 10 images, 5 videos, and 5 audio files. Cite them as Image 1, Video 1, and Audio 1, and say what each one controls.
Edit or extend a clip you already have
Edit rewrites one element and keeps the rest. Extend continues time forward, backward, or both ways. The original plus the new seconds must stay within 30 seconds.
Lock the first frame, or the first and the last
Frame mode generates the move between stills you already approved. It does not share a request with reference videos or audio. Send one mode per generation.
Standard or Prime
Both modes share the controls. Prime is the faster setting, not a quality upgrade. The picker is the rate card: Standard is 6, 12, and 24 K-coins per second at 480P, 720P, and 1080P. Prime is 9, 18, and 36. Start on Standard at 720P while the prompt is still changing.
Product ads, 9:16 social spots, two- or three-shot stories, and a line you can check by ear are the jobs this range fits. Approve the clip with sound on, not from the poster frame.
Wan 3.0 examples on KreadoAI
Four jobs, four clips. The prompts for repeating them are in the tutorial.
Example 1 — An 18-second watch film
Four beats inside the 30-second ceiling: a frozen wave, a move into the mechanism, the wave released, and the watch on volcanic rock. This pass was generated without audio so the picture could be judged on its own. 720P, 16:9.
18-second multi-shot watch film, audio off.
Example 2 — A cafe line with sound
A 10-second pour and a short spoken line over cafe noise. Listen once. A good frame with the wrong line is still a failed take.
10-second cafe pour with a spoken line, audio on.
Example 3 — A steak spot from four references
Image 1 is the plated ribeye. Image 2 is the dining room. Video 1 is the sear. Audio 1 is the sizzle. The prompt assigns each file one job.
10-second reference clip. No first or last frame was attached.
Example 4 — Edit the light, then extend the shot
The edit starts from a daytime street and asks for the same street at night in rain. The layout and the camera stay.
Daytime source, then the same street in night rain.
The extend keeps a short coastal aerial and asks the camera to keep flying over the same shore. A successful extend looks like more of that shot.
Coastal source, then the forward extend.
First-and-last-frame control is a different upload. The car clip uses a garage still and a coastal still, with no reference video attached.
8-second move from the garage still to the coast still.
How to use Wan 3.0 on KreadoAI
- Open the Wan 3.0 AI Video Generator and select Wan 3.0 Standard while the prompt is still changing. Prime is the faster setting once the prompt is stable.
- Pick one mode. Text, references, edit, extend, and first-and-last frames are different requests. Do not mix frame stills with reference videos.
- Set 16:9 or 9:16, 720P, and a length from 2 to 30 seconds. Turn audio on only when the prompt contains a line or an effect you want in the file.
- Cite every upload as Image 1, Video 1, or Audio 1, and name the one thing it controls.
- Watch the clip with sound, change the line that failed, and generate again. Paste-ready prompts are in learn how to use Wan 3.0.
FAQ
What is Wan 3.0?
Wan 3.0 is Alibaba's video model on KreadoAI for generating, editing, and extending clips up to 30 seconds, with optional native audio and mixed image, video, and audio references.
What are the Wan 3.0 features that change the job?
You can time a clip up to 30 seconds, put dialogue and effects in the same file, combine up to 20 reference files, edit or extend a clip you already have, and lock first and last frames in a separate request.
How is Wan 3.0 different from Wan 2.7?
Use Wan 3.0 for the timed clip, the native voice, the mixed references, and directional extension. Stay on Wan 2.7 when the job is the 9-grid workflow on that page.
Does Wan 3.0 Prime look better than Standard?
No. Prime is the faster setting. It uses the same controls and costs more K-coins per second. Switch after the prompt is stable.
Can one request mix first-and-last frames with reference videos?
No. Frame mode and reference mode take different uploads. Send one mode per generation.
New timed clips with sound start on Wan 3.0. A batch that already depends on the Wan 2.7 9-grid stays on that page.




