Inspiration
Christmas is coming soon, I want to create a story filled with hope, courage, and warmth.
What it does
I hope this song can replace Mariah Carey. ...I'm kidding, I know her status is unshakeable, and I also know my own work is very immature. In fact, the song I wanted to make was originally supposed to be over three minutes long. I wrote a complete story, and the MV content was supposed to be divided into three storylines. One line is Santa Claus getting ready for work (accompanied by a cute, chubby orange cat patrolling the factory, creating fun moments with the elves). The other two lines are set in the 1950s and the 2000s, respectively. In the snowy scene outside, a roof with a chimney under a shooting star symbolizes the 1950s, while a roof without a chimney symbolizes the 2000s. The realistic-style animation is the 2000s, featuring a man with a birthmark on his neck, who is happily decorating the house with his daughter. The cute-style animation is the 1950s, where a little boy is ostracized and has a hard time in the MV, but still encounters kind people who help him, filling him with hope and courage. The story is revealed at the end when the little boy takes off his scarf, showing that he also has a birthmark on his neck; he is the man's childhood self. I hoped to convey through this story that Christmas is a festival full of warmth, bringing hope and love to people, and that it is passed down this way. But unfortunately, I had too little preparation time, not enough tools, and very little money, so I could only use free resources. As a result, I encountered great difficulties in post-production. I had to cut the original song down to just over two minutes, and in the end, it only presents a very ordinary and plain Christmas.
How we built it
First, I would use Perplexity (Claude Sonnet, Gemini) to discuss the lyrics and style, then use SUNO to generate a few songs, and finally select the one I'm most satisfied with. After the song is generated, I will use Nano Banana to create images, use MeituPic and Canva for photo editing, then use grok to generate video, sometimes I also use flow to generate video, and finally use cupcut for editing.
Challenges we ran into
I think the hardest part is composition. I have the scene I want to shoot in my mind, but the AI-generated images often look different. I have to go back and ask Perplexity how to give instructions to the AI, then paste it into banana. Sometimes, one scene can take me an entire morning. Secondly, because I wanted to remove a piece of clothing for the composition, and since I don't know how to use Photoshop, I could only give commands to the AI to remove the little boy's scarf (because I needed to reveal the birthmark on his neck). However, the scarf was very difficult to remove. When I finally managed to remove it, the yellow coat turned out to be a turtleneck! The birthmark was completely covered, and then the yellow coat couldn't be removed. My free quota reached its limit, and I didn't have enough time, so I had to give up for now...
Accomplishments that we're proud of
Actually, I made another song featuring characters from Animal Crossing. The content explains that Taiwanese people are as friendly as the island residents in Animal Crossing. That song is very cute and successful, but I failed to get authorization from Nintendo after writing to them, so I can't submit it. The current song was made urgently in a day and a half. I have done my best on both songs.
What we learned
I think AI technology is improving more and more, but how to give instructions is really important. As I spend more time making these two songs, my proficiency in using AI also increases, and I feel I have learned a lot.
What's next for Music video submission
I will continue to make songs I like and share them on YouTube.
Log in or sign up for Devpost to join the conversation.