SuperAPI AI Directory

Qingying 2.0 Major Update, AI Image-to-Video Results Tested and Stunning!

Author: AI小集 · Original source: AI工具集

Qingying 2.0 was updated today. As the earliest product in China to fully launch to consumers, the hard power of Zhipu Qingyan is obvious to all.

Overall, after taking a look atQingying 2.0's results, the model quality is too stunning. AI video seems to have reached a new peak again.

01

Comparison of Qingying 2.0 and 1.0

Image quality, color, and motion are all greatly improved

Same as the previous version, you only need good ideas, type a few lines of text, add a little patience, and you can generate 4K high-quality videos.

Under the same image and prompt, let's take a look at the comparison of the generation results of Qingying 1.0 and 2.0:

Qingying 1 (left) and Qingying 2 (right)

For the most basic image-to-video,Qingying 2.0's overall image quality is higher, the colors are more vivid, and even the generated motion range is larger and more natural, what a versatile warrior!

Prompt: The parachute opens to the maximum, the camera rises, until the top of the parachute is visible

Qingying 1 (left) and Qingying 2 (right)

Qingying 2.0 can better understand and execute the instructions I give, and the new content introduced is highly consistent with the style of the original image.

02

Qingying 2.0 Test Video

Comprehensive ability ranks among the T1 tier

When I need to create a video with rich camera movement and delicate performance, itsimage stability and motion coherence, I can only say it's awesome.

Prompt: The camera follows the man standing up, adjusting his clothes, close-up shot

What's even more impressive is that the sound of foreign-language conversation accompanying the man getting up in the video was automatically generated by Qingying by understanding the video images!

For making AI videos in the future, there is no need to search everywhere for sound effects,the related voices and dubbing can all be automatically generated with one click, this effect is so surprising!

2.0's subject consistency has also reached a new level。

For example, in this case, the little elf dances in the forest, and the overall picture is very lively.

Prompt: The little elf vibrates its wings and dances in the forest

And it's not only like this when making realistic-style videos. When creating 2D animation and 3D animation, Qingying 2.0's expressiveness is also quite good.

The artistic style and visual beauty are directly maxed out.

Prompt: The kitten approaches holding a candied hawthorn stick, raises the candied hawthorn stick to the camera

However, these are not what I think is the most awesome part of Qingying.

Speed, and price, these are the key points that many people have not noticed.

For non-members, generating one video with Qingying 2.0 only requires waiting in line for a few minutes, and it is still free unlimited generation!

Cheap, fast, qualityall three balanced! At this moment, Qingying 2.0 is the most awesome AI video model, bar none.

03

More Case Demonstrations

1. Sports

Prompt: Make the picture move

When I do not specify a specific subject, it can also automatically recognize the subject and make movements that conform to physical laws.

Suitable for beginners to use, very easy to get started!

2. Sci-Fi Special Effects

Prompt: A spaceship flies over a space base

The overall style of the generated video is very unified, the motion is smooth, there is no deformation, and almost no multiple generations are needed.

3. Art Style

Prompt: Make the picture move

The model has a thorough understanding of images and color spaces, and it is also quite good for making abstract art videos.

IV. Realism

Prompt: The hot pot is boiling, the water is churning, and waves of steam are rising

AI's replication of real-world scenes has received an epic-level boost.

In the video, the scene of the water boiling and the steam both conform very well to the laws of physics.

V. Multi-person scenes

Prompt: All the women in the scene stand up and dance

In multi-person scenes, many video models tend to collapse, but Qingying 2.0 remains very stable! Everyone's movements look very coordinated.

04

Summary

Qingying's AI video capabilities have been comprehensively upgraded and entered the 2.0 era.

1. The base model is more powerful: comprehensive optimization of the model structure, training methods, and data engineering makes the videos generated by Qingying 2.0 smoother and richer in detail.

2. More stable results: the generated videos are more controllable, and even if the main subject in the frame moves with a large amplitude, the image remains stable. This reduces the time cost of repeated generation.

3. Improved artistic aesthetics: it can handle various artistic styles, and the visual aesthetics are greatly improved.

Most importantly,Qingying 2.0is completelyfree, and cangenerate an unlimited number of times!

As an all-around player among AI large models, Zhipu, while balancing both toB and toC businesses, can also rapidly push "Qingying" into the T1 tier, and may even become fully open source in the near future. It is undoubtedly a true pride of domestic AI!

In this pre-New Year update, Qingying also hid a small Easter egg. On the video generation page, selecting the gold coin effect can make your New Year full of wealth.

Finally, I wish everyone a Happy New Year in advance, and may you make a fortune in the Year of the Snake~

All tools involved in this article:

Zhipu Qingying:https://ai-bot.cn/sites/59040.html

Original link:Qingying 2.0 is fully launched, and domestic AI video has reached a new peak!