Choose a category
3 models available
8 models available
3 models available
2 models available
4 models available
8 models available
Wan 3.0 Video
Write it, cast it with reference images, get up to 30 seconds with sound.
Creative Vision
Setting, subject, action, camera. With reference images, say what each is for: “@Image1 is the character, @Image2 the set.”
Video Parameters
Priced per second — 2 seconds at 480P is 4 credits, 30 seconds at 1080P is 220.
Each step up doubles the price. Draft at 480P, finish at 1080P.
Auto follows your reference images; with none, pick a ratio.
AI audio synced to the video. Does not change the price.
Cinema Preview
Generation History0 records
Wan 3.0 Video Generator
Alibaba's Wan family, version 3.0, on the standard tier. Write a prompt of up to 20,000 characters, optionally hand it up to 10 reference images and one audio clip, and it returns a clip of 2 to 30 seconds at 480P, 720P or 1080P with its own audio track. The prompt addresses your uploads by number — @Image1 for the character, @Image2 for the product — so the things you care about stay the same from run to run.
Two seconds at 480P is 4 credits — enough to see whether a prompt reads before you spend on a longer take
Why Choose Wan 3.0 for Video
Wan 3.0 is a prompt-first model that also takes references. With no uploads it is a text-to-video model with a very long prompt box; with them it becomes a way to keep the same character, product or set across a series of shots without describing it again every time.
Up to 10 References, Called by Number
Uploads are numbered in the order you add them and you address them in the prompt as @Image1 through @Image10. Say which one is the character, which is the product and which is the set; the model keeps each one in its role instead of blending them into one look.
Any Length From 2 to 30 Seconds
A continuous slider, billed per second. A two-second test and a thirty-second take are the same setting at different positions, and you never round up to a fixed tier you did not need.
480P, 720P or 1080P
Each step up doubles the price for the same length: five seconds is 9 credits at 480P, 20 at 720P and 40 at 1080P. Draft at the bottom, deliver at the top, and pay for the difference only once.
Native Audio, On by Default
The clip arrives with an AI-generated soundtrack — ambience, effects and voice where the prompt calls for them — at no change in price. Switch it off when you plan to score the clip yourself.
A Reference Audio Clip Sets the Rhythm
Attach one mp3 or wav of up to 15 seconds and call it @Audio1 in the prompt. Its music, tempo or voice shapes how the generated track and the motion move together, which is hard to get from words alone.
Standard or Prime: Start With the Evidence
Both tiers expose the same prompt, reference, length, resolution and audio controls; Prime costs about 1.6 times as much. We have not published a controlled same-input benchmark, so standard is the sensible default. If you test Prime, keep every setting and seed unchanged and compare the outputs yourself.
Show It Once, Then Reuse It
A reference image is a description you never have to type again. Upload the same character to every run in a series and the face, outfit and proportions carry over from shot to shot.
How to Use Wan 3.0 Video
Length and resolution multiply, so the order below matters: settle the prompt and the references at the cheap end of both dials before you move either of them.
Add References, or Skip Them
Up to 10 images in JPEG, PNG or WebP, 20MB each, plus one audio clip of up to 15 seconds if you have one. Add them in a deliberate order — the numbers in your prompt refer to that order. Leave the uploads empty and the page works as plain text to video.
Cast, Then Describe the Shot
"@Image1 is the presenter, @Image2 is the bottle on the counter, @Audio1 sets the tempo. She picks it up, turns to camera, the camera pushes in slowly." Chinese and English both work, and the box holds 20,000 characters, so the shot can be as specific as you like.
Test at 480P, 2 to 5 Seconds
4 credits for two seconds, 9 for five. That is enough to see whether the model understood which upload is which and whether the motion reads. Fix the prompt here, where a mistake is cheap.
Extend and Raise the Resolution
Push the slider out and switch to 720P or 1080P for the keeper. Set a seed if you want to be able to re-run the same settings and get the same clip. Generation takes a few minutes — longer for 30 seconds at 1080P — and the result lands in your history.
References Are Not a First Frame
None of your uploads becomes the opening frame — they are material the model draws on. If you want the clip to start from a specific image, use the Wan 3.0 Image to Video page instead; the two inputs cannot be combined in one run.
Wan 3.0 Video
FAQ
Reference images, the @Image syntax, reference audio, resolution, pricing and how this page differs from Wan 3.0 Image to Video and the Prime tier.
Wan is Alibaba's video generation family, and 3.0 is its current generation. This page runs the standard tier with a prompt and optional reference images and audio; you set length, resolution and aspect ratio, and we handle generation, credit accounting and result storage. FlowVeo3 is an independent platform and is not affiliated with Alibaba.
Between 4 and 220 credits, set by length and resolution. Five seconds is 9 credits at 480P, 20 at 720P and 40 at 1080P; the full thirty seconds is 110 credits at 720P and 220 at 1080P. Reference images and audio do not change the price, and neither does the audio toggle. The exact cost for your current settings is shown before you submit.
Uploads are numbered in the order you add them, and you refer to them in the prompt as @Image1, @Image2 and so on, up to @Image10. "@Image1 is the hero, @Image2 is the sneaker she wears" tells the model which upload governs which part of the shot. A reference audio clip is addressed the same way, as @Audio1.
No. A run takes either reference images or a first frame, not both. This page is the reference side; if you want the clip to open on a specific image — and optionally close on another — use Wan 3.0 Image to Video, which is the same model at the same price with a different input.
Wan 3.0 Video Prime is a separately priced upstream tier with the same public inputs and controls, at roughly 1.6 times the price per second. We do not promise a specific quality or speed advantage without a controlled comparison. Start with standard; if you want to evaluate Prime, repeat the same prompt, references, seed, length and resolution so the model tier is the only variable.
Yes. A native audio track is on by default and does not change the price; turn it off if you will add your own. To steer it, attach one mp3 or wav of up to 15 seconds and call it @Audio1 in the prompt — it sets rhythm, music or the voice the generated track follows.
Usually a few minutes, and longer for 30-second 1080P clips. You can leave the page; the result lands in your history and stays available for seven days, so download anything you want to keep. If the generation fails, the held credits are released back to your balance automatically.
Try Wan 3.0 Video
Two seconds at 480P costs 4 credits and tells you whether the prompt — and the casting — is worth extending.