CONSTELLATIONS

Inspiration

It was night. I was thinking about entering into the music category while looking at the stars. I found the night sky beautiful and reflected further on how incredible it was that the stars could connect even trhough so far away. Then I got the idea: connectiong through space and slowly but surely it evolved into what it is today: "Constellations". It is the very idea that we're all connected through sounds. Afterwards, I wanted to push it further by creating something with many languages and since the rules stated; # Other related genres including K-Pop, J-Pop, C-Pop, Latin-Pop, Afro-Pop, Dance-Pop are all eligible in this category # , I went for it. I wrote the lyrics myself first in english and then translated the different parts.

Lyrics

Intro (Korean): I search for you like stars in the night sky, You become the light on hidden paths.

Verse 1 (Spanish): You are the fire that never fades, Shadows dance, but you embrace me. Your voice guides me like a river to the sea, A destiny I cannot stop.

Pre-Chorus (Japanese): Escaping the maze of the heart, The future we draw together is infinite. Even if the riddle won’t be solved, With you, it becomes the answer.

Chorus (English): We’re constellations in the dark, Different skies but one same spark. Across the borders, hearts align, Your light is forever mine.

Verse 2 (Mandarin): In shattered dreams, I hear you, Like dawn igniting my soul. No matter the distance, we’ll draw near, Your name written across galaxies.

Rap (French): Free like the wind, I fly without fear, Your rhythm is my compass, it keeps me elsewhere. Energy in the air, we burn the sky, No ocean can stop the eternal.

Bridge (multilingual): The road to you has no end. Each step brings me closer. A melody that lasts forever. Our hearts breathe in sync.

Final Chorus (English): We’re constellations in the dark, Different skies but one same spark. Across the borders, hearts align, Your light is forever mine.

How I built it

It took many steps and I worked for around a month on this one which was my biggest and longest music video project to date.

1. The Sound and Singers => First, I wrote the lyrics myself. => Then I decided on which language to keep for which parts. Translated and optimized using Gemini from Google. => Afterwards, I started to consider the number of singers. At first I wished to get one for every language but then I reconsidered. The song has six languages within and I wanted some stability kept to get a ryhtmic coherence strong enough. I settled on three with a main female singer and two supporting male singers, with a similar lyric repartition in 1:2. => I then went on to craft the song using ProducerAI/Riffusion. It took a while, many different generations just to get the background music and then to get the perfect singers's voices (and at the right placement) => My idea was a korean female singer from the start (since in K-pop, they usually are able to sing Korean but also Japanese and English) and then I was left with the Spanish, Chinese and French parts. I decided to get a single singer for the Chinese/French part and went from there. At first I had considered two female singers but I simply 'fell in love' with the voice I managed to generate for the first part and, well, I wanted to hear more of that one. Also, I decided to include the DJ within the characters to make it four, the Dj linking the voices with his sound. I actually used my own self for it so the DJ in the video is actually 'me' as in I used one of my pictures and through Google Nano/Imagen, crafted visuals of myself. => I moved on to crafting the different outfits for each of my characters. I used GPT (OpenAI) and Google (Imagen Nano / Gemini) to get ideas and craft the outfits. => I used Google (Imagen Nano / Gemini) to create the three singers: female korean, male spanish, male chinese and readapted the picture of myself.

2. the Video Clip

=> First I created some intro video with my own watermark "LIGHT" for a simple start as I wanted the video to clearly start with the DJ, with the one linking the voices. He turns on the set, he begins the music. => Then I went on to considering what to do. The configuration, by this point, felt obvious: they are singing in different places. Places filled with history so that it gives a value in 'time' to it. For Korea, it's Gwanghwamun Gate, Spain is the Sagrada Familia, Tokyo Tower for Tokyo, The Chinese Great Wall, The Eiffel Tower in Paris, Gaza Pyramids in Egypt and finally back to the Colosseum of Roma in Italy. I picked the Colosseum for the first and last parts because I went there once and found it incredible, probably my favorite historical monument hence why I wanted the first and final links to happen there. => I used Google Imagen Nano to create the First and Last frames and Hailuo 02 FF/LF to craft every single part of the video (6 or 10 seconds at a time for the rendering and then cutting/editing in VSDC) and so on. It's hundreds of video generations to get the 'perfect' ones. => I was constantly on VSDC as I generated and edit at the same time so that I could keep a perfect view and continuity as well as coherence between the singing parts (location-wise, singer-wise) and had to cut and redo many parts, many times. => Every time I finished a singing segment (korean / spanish / Japanese / English / Chinese / French / Multilingual in Egypt / Final in Roma), I directly went to Higgsfield (using KLINGAI) to fix the lipsync before moving on to crafting the 'transition' and then the next segment. => Finally, I added the ChromaAwards logo at the start and end of the Music Video.

Challenges I ran into

There were many difficulties but I really enjoyed working on this project because it truly felt 'mine' in a special way. I loved the lyrics I built, I loved the song itself and then, well, I made everything to the best I could in a way I loved. The biggest difficulty was to create and generate the song perfectly. To get different singers with perfect clear voices and at the right timings was not easy. To say the least, it is what took me the longest time. I'm glad I took the time for it considering the result though. Beyond that, the hardest shot to make was for the chinese man in Paris when he launched his beret to create a transition towards Egypt. To configure the right prompt and get that one as I had envisionned it felt incredible. I got very excited to succeed. Aside from that, it was also quite difficult to craft the ending in Roma. To successfully add in the different characters to join in and interact correctly forced me to adapt the prompts little by little for quite a long time too.

Accomplishments that I am proud of

There are a few but the biggest one is simple: I made a music video I am proud of Nothing can compare to that.

Everything else consists of small achievements within:

• I managed to add my own self into the song, interacting with my virtual singers.
• I was able to craft a perfect transition with the beret in Paris.
• I made a song I simply love.

What we learned

I did learn a lot during the project:

• I was able to work further in understanding the coherence of lyrics, voices and sound in a multilingual style.
• I was able to consider fashion outfit designs, locations and visuals that all go together to link together.
• I was able to truly use the lip sync to its potential.
• I was able to create transitions between locations and individuals beyond a simple 'cut to' and truly link the entire music video together.
• I was able to include my real self into my visual creation in a way that felt natural.

And, more than anything, I fell in love again with music and AI again and even deeper.

What's next for CONSTELLATIONS

I will be listening to it, others too. I will send it to some DJ friends in Paris and in Russia to potentially play in clubs or festivals. It was a great step for me in learning how to create even better music and videos so I will remember it forever regardless. And, who knows? Maybe one day I'll be playing it myself in the Colosseum.

Built With

• gemini
• googlenano
• gpt
• hailuoai
• higgsfield
• klingai
• producerai
• riffusion