Vidu AI Video Generation, In Your Browser
Two Vidu generations sit in the same studio as everything else. Here is which one to pick, and what each is actually good at.
Vidu comes from Shengshu Technology, a lab founded in 2023 with its core team out of Tsinghua University’s AI institute. It went public in July 2024, and by the time Q2 arrived in September 2025 its reputation rested on one claim more than any other: keeping a subject recognisable when you generate the same character or product across several shots.
OpenFilmo runs Vidu Q1 and Q2 inside a single browser studio rather than as a separate subscription. You choose the model in the generator panel, set an aspect ratio and an output resolution, and pay the same credit rate as every other tool in the workspace — there is no per-model surcharge and no second account to manage.
What it does
Built for this specific job, not a generic generator.
Two generations, one balance
Q2 is the newer model and the default choice on anything with a face in it. Q1 stays selectable when you want the flatter, more literal movement of the earlier generation without leaving the workspace.
Reference conditioning
Supply images of the subject and describe only motion and camera for each shot. This is what makes a multi-shot sequence usable rather than a set of clips that merely resemble each other.
Camera and aspect control
Push-ins, pull-backs and tracking shots are describable rather than left to chance, and framing goes from vertical social crops through 16:9 to wider ratios without re-uploading anything.
Shared credits with the rest of the studio
A clip, an image and a 4K upscale all draw from one balance, so a finished piece does not need three separate quotas or three separate invoices.
How it works
Four steps, no timeline to learn.
Prepare references first
One clean, evenly lit image of the subject beats five inconsistent ones. Any variation you feed the model is variation it will reproduce later.
Pick the model and the mode
Q2 for performance and expression, Q1 when you want the earlier look. Choose the faster mode while you are still testing the prompt and switch once the composition holds.
Describe motion, not appearance
Appearance is already settled by the reference. Spend the prompt on what changes between frames — the action, the camera move, the beat in the sequence.
Check at 100 percent, then extend
Faces and edges are where problems hide. Approve the short test before extending, because extension compounds artefacts rather than resolving them.
What people make with it
The jobs this actually gets used for.
Short-form drama and serials
The same character across a run of short scenes is exactly the problem reference conditioning was built for, and it is where a consistent face saves the most reshooting.
Product and brand ads
Hold one product or presenter steady across several cuts so the ad reads as one production rather than a montage of near-misses.
Talking-character content
Explainers and avatar-led clips where the value is in facial detail and believable delivery rather than spectacle.
Catalogue and social variants
Generate the same subject in several framings and lengths so one asset stretches across placements instead of one take serving one channel.
Why the model choice decides the outcome
Most disappointing AI video is not a rendering failure. It is a consistency failure.
A clip can look excellent in isolation and still be useless, because the character in shot three does not match shot one and no amount of grading fixes a different face. That is why the useful question is not which model produces the prettiest frame but which one keeps a subject stable long enough to build a sequence.
Reference conditioning is the mechanism that addresses this, and it is a conditional advantage: it works when the reference images are clean and consistent, and it degrades when they are not. Preparing two good images is worth more than any prompt trick applied afterwards.
It also pays to separate the two jobs a model does. Appearance comes from your references, motion comes from your prompt. When those are mixed into one instruction you lose the ability to fix one without disturbing the other.
Credits and plans
Generation is metered in credits, shared across every tool.
A five-second video costs 5 credits and one image costs 2. The Free plan gives you 30 credits every month with no credit card, no watermark and commercial use included. Paid plans start at $9 for 400 credits and add faster generation, parallel tasks and higher export resolution.
Related guides
Longer reads on the same workflow, from the OpenFilmo blog.
How to Use Vidu: A Practical Guide
How to use Vidu: reference-to-video workflow, keeping subjects consistent across shots, prompt approach, and where it fits compare…
Character Consistency in AI Video: What Actually Works
How to keep characters consistent across AI video shots: reference images, prompt discipline, key frame locking and post-hoc corre…
Text to Video AI: Turn Any Prompt Into Video
Learn how text to video AI works, the four-part prompt formula that produces usable clips, how to fix common failures, and where i…
Related workflows
Closely related tools and guides on the same problem.
Frequently asked questions
Which Vidu models can I use here?
Q1 and Q2. Both appear in the generator panel and you switch between them per generation, so you can run the same prompt through each and compare rather than committing to one.
What does a clip cost?
Five credits per five-second clip. The Free plan gives 30 credits every month with no credit card and no watermark, which is about six five-second tests before you pay anything. Paid plans start at $9 for 400 credits.
Do I need to install anything?
No. Generation runs in the browser from the same workspace as the image and upscaling tools, so there is no desktop app and no separate Vidu account.
Why is reference conditioning so important?
Because multi-shot work fails on consistency far more often than on quality. Supplying images of the subject lets you describe only what changes between frames, which is the part you actually control.
Can I use the output commercially?
Yes. Commercial use is included on the Free plan as well as on paid plans, and Free output carries no watermark.