-
-
Thunmbnail
-
Scene #1 (The platform)
-
Scene #2(Recall)
-
Scene #3(The street)
-
Scene #4(In the train 1)
-
Scene #5(In the train 2)
-
Scene #6(Fallen leaves road)
-
Scene #7(Sorrow)
-
Scene #8(Flashback)
-
Scene #9(Desolation)
-
Scene #10(Fallen leaves)
-
Scene #11(fallen leaves2)
-
Scene #12(recall 2)
-
Scene #13(Platform 1)
-
Scene #14(Late autumn 1)
-
Scene #15(Late Autumn 2)
Inspiration
'Late Autumn', a K-POP ballad song that contains the longing for the sad season that remains in the frozen time, I wanted to express it in a video. My goal was to visually realize this emotional theme using the latest generative AI technology and maximize production efficiency.
What it does
This project is the result of producing the official music video for the K-POP ballad 'Late Autumn.' It actively utilized AI tools to visually embody the song's feeling of coldness and melancholy (哀傷) and validated an AI-based End-to-End production pipeline capable of quickly completing high-quality content with minimal human resources.
How we built it
The production process was structured as a sequential combination of four key AI tools:
Music Generation (Suno): Generated the K-POP ballad music and vocals from scratch to match the song's atmosphere and lyrics.
Video Image Scene (ImageFX): Produced high-resolution still images that aligned with the song's lyrical and atmospheric mood, preparing the core visual elements.
Video Clip Generation (Kling): Converted the core images created by ImageFX into short video clips, securing the moving scenes required for the final music video.
Video Editing and Finalization (Capcut): Combined the audio generated by Suno with the video clips generated by Kling. The final music video was completed after precise synchronization work on clip transitions, effects, and the soundtrack.
Challenges we ran into
The biggest challenge was maintaining visual consistency when connecting multiple AI tools. It was difficult to unify the subtle differences in style and color between the scenes created by ImageFX and those animated by Kling. Additionally, the Temporal Consistency of the video clips generated by Kling was imperfect, requiring meticulous manual adjustment at the frame level in Capcut.
Accomplishments that we're proud of
We successfully established and validated a complete music video production pipeline using only AI tools from music creation to final editing. We succeeded in visually expressing the sad and longing emotions of 'Late Autumn' without complex manual work, proving the creative expressive capability of generative AI. This pipeline significantly shortened the production period.
What we learned
We confirmed that unified guidance in the initial prompt is key to determining the consistency of the entire pipeline when using multiple generative AI models (Suno, ImageFX, Kling) in sequence. Specifically, we learned the importance of a hybrid workflow that compensates for the limitations of Kling's video output using a professional editing tool like Capcut.
What's next for Late Autumn(Music Video)
Building on this successful AI production pipeline, we plan to immediately apply it to the production of music videos for other songs. Furthermore, we aim to further enhance post-production efficiency by incorporating viewer feedback and developing a Capcut automation script to complement Kling's output.
Built With
- capcut
- imagefx
- kling
- suno
