Domestic large model manufacturers are becoming increasingly competitive in the multimodal direction. First there was the overseas sensationKling, and then the breakoutMiniMaxHailuo. According to Tencent Hunyuan's official announcement, on December 3 it will officially release theHunyuan text-to-videolarge model.
Kimi, which has been continuously shrinking its business lines and focusing on core products, recently also quietlygray-testedthe "Kimi Creation Space" feature
You only need to enter one sentence + one piece of music, and it can generate a music video up to 30s long。
Official cases
First, let's enjoy the official video cases:
1. In the Mood for Love style
2. The Grand Budapest Hotel style (beat-synced)
3. Cyberpunk style
Don't they all look very impressive? Whether it'slightingandcolor, they are all very professional, and some also usebeat-syncedediting methods,the generated videos can be directlyused, lowering yet another barrier forzero-foundationusers to make videos with AI.
"Kimi Creation Space" hands-on test
- Test 1 MV production
A few months ago, MiniMax's Hailuo AI launched the feature of AI-generated music. Now that Kimi has launched the feature of generating music videos, 🤩 does that mean ordinary people can also make music and produce MVs at low cost?
For example, for this song "Clouds of Hometown", I input "beautiful scenery of hometown, grassland, white clouds, mountains and forests" to see the images generated by Kimi:
It will split the music into multiple clips according to the rhythm. For each clip, it will also write a set of keywords for you and generate the corresponding video images.
If you are not satisfied with a generated clip, just edit the description of that single clip and generate it again. 👍 The overall style consistency of Kimi's images, and the level of refinement of the images, are something other software has never achieved before.
- Test 2 Daily short video production
How to write keywords has always been a difficulty for beginners learning AI from zero foundation😭, but now, you only need to input your idea, without writing keywords, without editing or secondary processing, and Kimi can help you complete an AI video, with quite good quality.
For example, I input: "A cat playing with a ball in the living room", and the images it generated look like this:
The overall image is very beautiful, but there will still be some flaws in details/close-ups.
- Test 3 Short drama production
A bit more advanced, what if we incorporate a little plot into the music?
Using Kimi's text feature, this idea can be easily realized.
Using the description Kimi gave me to generate a video, the result is like this:
What might usually require a combination of several AI tools and repeated generation attempts, Kimi can reach a similar level in just a few minutes.
The generated video has a very unified style and color tone from beginning to end, the plot perfectly matches the description I input, and the character's appearanceconsistency is maintained well, even details such as clothing and backpacks are the same.
- Test 4 Long plot production
If the plot is a bit longer, how will Kimi perform?
At present, Kimican generate a video up to 30 seconds long in a single pass. We entered "couple breakup", and the generated video looks like this:
Judging from the videos generated in the actual test, the character presentation is also very coordinated overall, and the characters' movements and expressions are quite smooth, so it does not look like there will be auncanny valley effect；
Kimi's new feature is currently still in grayscale internal testing, and friends who have not been included in the grayscale rollout need to wait a bit longer.
Currently, 100 seconds of video can be generated for free every day. The experience duration does not accumulate and refreshes daily.
Actual test summary
This new feature is already very top-tier in terms of understanding and generating content👏,as long as you can type, you can generate fairly good-quality music videos, but it is still difficult to apply to work that requires fine-grained control of details. After a few more iterations, it can be applied to advertising (creative previews)🎬, film (shot effect previews)🎥, and games (special effects generation)🎮.
Using Kimi's music video generation feature to create MVs 🎵, travel logs ✈️, anime, game CG 🎮, personalized e-cards💌, event invitations🎉... is all very convenient.
Kimi has already become my work and study partner. Summarizing and distilling content, searching for information online, reasoning through problems, etc. are real productivity. This attempt at AI creative creation features truly amazed me😍. I look forward to Kimi continuing to launch more fun and interesting features!
All tools involved in this article:
Kimi：https://ai-bot.cn/sites/5355.html
Original link:Get an early experience of Kimi's video generation feature, it amazed me!