Recently, the AI image production track has been going crazy!
From the explosive popularity of Nano Banana to the continuous launch of Zhimeng AI 4.0 and Doubao 4.0, Vidu, which has always focused on video big models, can’t help but:
**Vidu Q1 reference image * * makes a grand appearance! Simultaneously supports * * 7 * * references.
Subject consistency is no less impressive than Google Nano Banana.
(Nano Banana supports up to 3 reference images)
Quantum bits took the lead in testing this model, and its performance was quite good -7 reference images that can be freely referenced bring extremely high operability.
Simply describe it in natural language.
Alternatively, it can directly generate fashion blockbusters, saving on on-site filming and other related tasks.
We have explored many interesting gameplay options, including prompts and pictures below. Let’s take a look together!
##How can I play with 7 reference pictures?
We tested several gameplay methods, such as combining various discordant elements into a harmonious image, or creating fashion blockbusters
It can be said that as long as there is creativity, everything can be synthesized.
####Everything can be synthesized
Whether it’s letting Qin Shihuang ride a polar bear and drink lemon water in Shanghai:
Let Li Bai successfully land on the moon by rocket:
As soon as the reference image is displayed, let’s see if the results of Vidu Q1’s reference image match our imagination.
####Trendy items instantly become OOTD
Since there are so many references, wouldn’t it be possible to directly replace the entire set?
All items can be worn with just one click, it’s time to showcase the power of matching (get ready).
Using this set of prompts is feasible for both the Mediterranean and the West Coast.
Online fitting is no longer a problem.
####One person can also shoot fashion blockbusters
Just changing clothes is not enough. Going further, how about generating a poster level photo?
Look, with simple prompts, the background, lighting, and composition can all be adjusted freely.
####A group photo? No, it’s a group photo
A group photo of two people is no longer enough to meet the demand, would you like some more?
How about gathering the young anime protagonists together.
Even if there is a significant difference in art style, even across a dimension, it can naturally appear in the same picture.
##More gameplay waiting to be explored
Of course, there are many more interesting gameplay options.
Basically, what Nano Banana can do, Vidu Q1 can also refer to the production chart.
Whether it’s the handmade creation of Xiao’s popular book.
Or the design draft should be transformed into a 3D stereoscopic effect.
Or perhaps the most classic subject replacement.
Most similar products only support 1-3 reference images, and many cannot support multi-element references of the same type, such as referencing multiple characters simultaneously.
The Vidu Q1 reference image can support 7 reference images at the same time, and still ensure that the appearance and features of the characters/subjects do not deviate in multi person, multi scene, and multiple generation scenarios.
The test data comprehensively surpasses Flux Kontext and directly benchmarks Nano Banana.
The consistency ability score is the best in the whole game.
Finally, let’s briefly introduce the usage of Vidu Q1 reference image.
On the reference image page, it supports 7 reference images, which will be displayed in the input bar below after uploading. You can drag and change the prompt words.
You can also enter [@] to select an image.
Clicking on ‘Create Subject’ allows you to input your commonly used data, and perhaps the three view effect would be better.
At present, there is only 1080p resolution, and the aspect ratio and number of generated images can be selected, with a maximum of 4 images generated at once.
Vidu Q1 reference image is now globally synchronized and online. Welcome to share your works in the comment section
*Actual testing address: https://www.vidu.cn/create/reference2image *