Creating a song with AI takes minutes, but creating one you actually want to release takes a little method. This guide walks through the whole process as it works in MixingMusic, which uses the Suno V6 model through KIE.ai: from the first idea to a mastered file.
1. Start with one clear idea
Before you type anything, decide three things: genre, mood and what the song is about. "Romantic bachata about a long-distance relationship" gives the model far more to work with than "a nice song". You do not need music theory; you need a direction.
2. Write the style (the prompt)
The style field describes how the song should sound, not what it says. Keep it to short, comma-separated descriptors:
- Genre and sub-genre: latin pop, reggaeton, acoustic ballad
- Instruments: nylon guitar, 808 bass, live drums, strings
- Mood and energy: uplifting, melancholic, high energy
- Voice: choose female or male vocal in the form, or describe it
Avoid instructions such as "follow my melody" or "make it sound like [famous artist]". The first is not something the style field controls, and the second can get the song rejected.
3. Lyrics: write them or let the AI do it
You have two paths:
- Paste your own lyrics. Use section tags such as [Verse], [Chorus] and [Bridge] on their own lines so the model knows where each part goes.
- Leave the lyrics empty. MixingMusic writes original lyrics from your description before sending the song.
Your own lyrics give you control over the message; AI lyrics are a fast way to get a first draft you can then edit and regenerate.
4. Optional: start from your voice
If you have a melody in your head, record it or hum it in the browser. Remix keeps your melody and timing and dresses it with the style you wrote. New in its style uses the recording as inspiration for a different song.
5. Choose the length and generate
Songs run from 1:00 to 4:00 (up to 2:30 on the free plan). Each generation uses 1 song from your balance. If the engine rejects the song, it is not deducted.
6. Compare the two versions
Suno V6 returns two versions of every song. Listen to both in the player with the Version 1 / Version 2 buttons. Pay attention to the chorus, the clarity of the lyrics and how the song ends; those are the points where the versions usually differ most.
7. Split the stems
Open the song menu and export the stems: vocals, drums, bass, guitar, piano and other, as WAV files. Stems are what turn a generated song into something you can actually fix.
8. Mix and master
Open the stems in MixingStudio to rebalance the vocal, tame the low end or brighten a dull instrument. Then master the result for streaming. If you prefer, you can skip the stems and master the stereo song directly.
Common mistakes to avoid
- Writing a paragraph of instructions in the style field instead of descriptors.
- Using lyrics from an existing song: the copyright filter will block it.
- Judging the song only on laptop speakers; check it on headphones too.
- Forgetting to download: songs are kept for 30 days and stems for 7.
Try it in MixingMusic
Create your song with the Suno V6 engine: 1 free song up to 2:30 with its stems, two versions of every song, and mixing and mastering in the same account. Songs and recordings are kept for 30 days and stems for 7 days, so download what you want to keep.
MixingMusic.AI uses the Suno V6 model through KIE.ai. MixingMusic.AI is not affiliated with or endorsed by Suno; Suno is a trademark of its owner.