How to Make Trot Songs With Suno AI
Here’s how to make Korean trot songs with Suno AI. I upload AI-made trot to my channel Eolssigu Trot every week. If Sylvamoor is my no-lyrics BGM channel, this is the opposite — songs with words, in Korean, which is a whole other difficulty. A plausible melody isn’t enough; the lyrics have to feel like trot and the chorus has to hook you. Tools: Claude + Suno + Higgsfield + CapCut.
Quick context: trot is Korea’s classic retro pop genre — think emotional ballads and up-tempo singalongs, beloved by older audiences and having a big revival. (“Eolssigu” is a Korean cry of joy, like “hurray!”)
A recent video:
Step 1 — Set the concept in one line
Everything starts with a one-line brief: what mood, and what singer persona. For example: “high-energy K-trot, middle-aged male singer, powerful vocals” or “sorrowful ballad trot, middle-aged female singer, clear and mournful voice.” That single line steers all ten songs of the day.
Step 2 — Write titles & lyrics with Claude
Given the brief, Claude first generates ten trot titles (same mood, rotating the emotional angle: love, hometown, life, the night stage…). Approve the titles, and it writes full lyrics for each.
The lyrics follow rules — and this is what makes trot sound like trot:
- Verse–pre-chorus–chorus structure
- An eight-line chorus, ten syllables or fewer per line
- A hook phrase from the title, repeated two or three times — but never in the same position twice
Along with lyrics, each song gets Suno style tags. Keep the singer persona fixed across all ten songs and rotate only tempo, lead instrument, and mood — that’s what gives the set album-like cohesion.
Step 3 — Generate the songs with Suno
Title + lyrics + style tags go into Suno, and out comes a song. It handles Korean lyrics surprisingly well; when pronunciation gets mushy, reroll.
Step 4 — Thumbnail with Higgsfield
This channel’s thumbnails have a quirk: the day’s singer persona appears in them. The singer from the brief — say, a middle-aged man in a sparkly stage suit — gets rendered by Higgsfield, with all ten titles overlaid in Korean. One thumbnail doubles as the video background.
Step 5 — Assemble in CapCut
Stitch the ten songs, lay the thumbnail as the backdrop, done. The description gets a timestamped tracklist.
Automate it + try it yourself
This whole workflow is packaged as a Claude skill too. One line of brief → ten titles → (approve) → lyrics, style tags, thumbnail, YouTube title and description, all as one set. Most of the creative labor is compressed into a single decision: “what mood today?”
You can play with the front half of the recipe (title + chorus) yourself — I built it into the 🎰 Trot Vending Machine.
The result
Two subscribers, three videos — very early. But trot is a genre where views actually happen: older Korean viewers leave long medley videos playing for hours. Whether AI-made trot works in that market is this channel’s research question. Results logged here.