Friendship, aging, and the bittersweetness of farewell. 🎬 See how @AmirAigc brings this heartfelt story to life with Kling 4.0. https://x.com/i/article/2115745028328869888
Amiraigc: Kling 4.0 Brings Subtle Emotions to Life!
In Tea When You Have Time, @AmirAigc puts Kling 4.0 to the test with a heartfelt story of friendship, aging, and farewell. Here's how he brought the story to life.
Notes from the Creator
When Kling invited me to create a short film with its latest model, Kling 4.0, I was free to tell any story I wanted. But sometimes, the more freedom you have, the harder it is to know where to begin. After much thought, I decided to make Tea When You Have Time, a touching story inspired by what I wanted to express and the model's new capabilities.
The inspiration came from a post I happened to read about a group of elderly Chinese people living overseas who gathered for tea. There wasn't a single deliberately sentimental line in the post, yet it carried a quiet acceptance of life's unpredictable reunions and farewells. You never know if you'll get another chance to meet.
There were other stories I wanted to tell, but choosing Kling 4.0 for this film was no accident. I felt that Kling 3.0 had a special strength in conveying subtle emotions, and I had high hopes for 4.0. The best way to understand a model is to put it to the test.
After working with it, I found that Native 30s Generation, Omni Reference, and Multiple Keyframes gave me more room to shape the actions and pacing. Character consistency was more stable than before, and I was also pleasantly surprised by how well the Cantonese dialogue matched the lip movements.
Native 30s Generation: Finally, Room for a Scene to Breathe
Compared with Kling 3.0, which could generate clips of up to 15 seconds, Kling 4.0 can generate up to 30 seconds at a time. That extra time gives a scene more room to breathe. It is not simply a matter of stitching two shorter clips together.
With a 15-second limit, I often found it difficult to fit an entire performance into one clip. I would have to extend the video or take a frame from one generation and use it to create another clip. Either approach could affect consistency, efficiency, or image quality.
With 30 seconds, an entire sequence of actions can unfold in a single generation. I can show the room, then let an elderly character enter, pour tea, dry their hands, pick up the phone, and answer it. The actions can unfold at a pace that feels natural for an older character, while the setting remains consistent throughout the scene.
Omni Reference: Keeping 12 Characters Consistent
In Kling 4.0, Kling has brought its multi-reference features together under Omni Reference, making them easier to access and use. The model also supports a 21:9 aspect ratio for a more cinematic look. But what matters most is how well it understands the references and keeps them consistent.
Tea When You Have Time was a challenging film to make with AI. It features a range of characters, from elderly people and children to robots, with 12 character assets in total, not counting the props and settings. Keeping them all consistent was a real test for Kling 4.0.
In practice, I found character consistency much more reliable than in version 3.0. The elderly characters, children, and robots remained distinct, without their identities getting mixed up.
Multiple Keyframes: More Control Over Every Shot
Alongside Omni Reference, Kling 4.0 also offers Multiple Keyframes as a separate feature. While you can already use prompts to set keyframes within Omni Reference, having a dedicated feature makes the process much more straightforward.
When working with AI, it's not always easy to get keyframes to stay exactly where you want them. They can sometimes drift, especially when you're still learning the workflow or the model produces unexpected results. Multiple Keyframes helps make that process more reliable.
For scenes where I need tighter control over the storytelling or want to maintain a particular visual style, I sometimes prefer to create individual keyframes rather than rely entirely on multiple references. This gives me a clearer way to guide the scene toward what I have in mind.
For the final emotional shot in Tea When You Have Time, I created three images before generating the video: an elderly character for the opening frame, a POV shot of a hand wiping away tears for the middle frame, and a child for the final frame. I then used Multiple Keyframes to bring them together in one sequence.
Kling 4.0 now supports up to 10 keyframes, which is great news for anyone who likes having a little more control over their shots.
Audio and Lip-Sync: The Cantonese Is Better Than Mine
I'm from Guangdong, and honestly, the Cantonese in Tea When You Have Time is better than mine.
The natural-sounding dialogue also struck a chord with Cantonese-speaking viewers, including audiences in Hong Kong and Macao. I was genuinely surprised by how well the model handled Cantonese. Rather than say more, I'll let you hear it for yourself.
According to Kling's official information, the model also supports a range of other accents and dialects, including Beijing Mandarin, Taiwanese Mandarin, Northeastern Mandarin, and Sichuan dialect, and American, British, and Indian English.
A Final Toast to Technology!
I'd like to close with a comment from a viewer who watched the film: “Technology for good. That's something I can get behind.” It really stayed with me.
As AI filmmaking tools continue to improve, making films has become easier, and audiences are becoming more open to AI-generated content. But their expectations are growing, too. They're looking beyond whether something looks fake or overly polished, and paying more attention to the emotions, ideas, historical accuracy, and ethical questions behind the work.
I saw this in the reactions to Tea When You Have Time. Many viewers were moved by the story, while others looked closely at its historical details or questioned the accuracy of the toy models. I think that's a good thing. It means the story connected with people, and they cared enough to look closer.
Of course, the film has its shortcomings, and the technology still has room to improve. Telling a gentle, unhurried story about older people in Cantonese may not be the easiest way to capture attention in today's fast-paced world. But at its heart, it's about the memories, emotions, and human connections we all share.
There's still a lot to learn, and the tools will keep evolving. Even when easier paths are available, I believe some stories are worth taking the longer road to tell.
So here's to technology, and to everyone using it to do good.
Originally written in Chinese by Ami without AI assistance. Translated and edited for clarity.