Update @20240324:
Suno.ai v3 is now available to all users! V3 allows you to make full, 2 minute radio-quality songs. It’s a notch higher than v2, and AI vocals’ recognition of Chinese has been improved by a large margin.
This is a translated version of a Chinese article, the original is as follows: 我是怎样一步步用 Suno.ai 生成歌曲 demo 的？
A few days ago, I saw a case of generating songs with Suno.ai shared by a dude (@jesselaunz) on Twitter. So I wanted to give it a try myself.
Suno’s free account gets 50 credits per day and each generation costs 10 credits, so you can generate 5 times per day. Each time, two song versions are generated for you to choose from. Each song is about 1 minute 2 minutes long.
Auto Mode
The default generation mode is automatic. Just describe the music style and topic, you can also describe the genres and vibes, even keywords, etc. (like ChatGPT Prompts).
In auto mode, I generated two songs in English and Chinese:
English https://app.suno.ai/song/2b42532b-1065-4139-97c4-50511cc6d890
Chinese https://app.suno.ai/song/599d109a-2337-4fed-b1ed-4b97f99a03fe
Custom Mode
I want to use my own lyrics, so I chose Custom Mode. There are 3 fields to fill in:
- Lyrics
Two paragraphs of up to 8 lines work best. If you have too many lyrics, you won’t be able to generate all of them, because there is a limit to the length of the song. Regarding the format of the lyrics, I think there are two key parts that should not be missing: the Verse and the Chorus. Lyrics should be like this:
[Verse]
Snowflakes fall, the fire’s burning low
I’m sitting by the window, watching them go (watching them go)
The tree is adorned with lights so bright
But it don’t feel like Christmas without you here tonight (without you here tonight)
[Chorus]
I’m dreaming of a white Christmas, my love (ooh-yeah)
With every Christmas card I write (with every card I write)
May your days be merry and bright (bright, yeah)
[Verse] and [Chorus] in the lyrics are markers that AI can recognize, indicating which part of the song the subsequent lyrics belong to. Other markers are [Intro] [Bridge] [Outro] [Instrumental] and so on, in parentheses or brackets.
Regarding song components, this song, also generated by suno, is a typical example (or is it official instruction?)
https://app.suno.ai/song/6ba1d5ce-ffa4-41cd-8146-779904fb908a/
- Style of Music
Fill in the music style you want. For example heartfelt anthemic folk, or Chinese pop style, sad love songs, female singers.
- Title
This is easy, fill in the title of the song you want.
Once you have filled in all 3 fields, click the Create button and two songs will be generated. You can choose one you are more satisfied with.
Continue From This Song
After first generating, you can continue to generate. Click on … to the right of the song you choose and select Continue From This Song in the menu.
This operation is important! Because Continue From This Song allows you to continue the previous style. Otherwise, if you start over, you’ll have a different style of music and it won’t continue from the previous one.
After this, on the left side of the screen, there is an additional section Continue From.
Now it’s time to fill in the second song lyrics in the Lyrics field. The music style should be same with the first song, it doesn’t matter if the song title is same or not. Click on the Create button and the second song will be generated. For songs that continue from the previous song, the words Part 2 will appear.
As usual, the third song continues to be generated from the second song, and the fourth song continues to be generated from the third song …
Get Whole Song
After generating from the second song, you can get the whole song. Click on … to the right of the second (or third, fourth) song and select Get Whole Song in the menu.
In this way, a complete song is generated.
Summaries
- Free accounts have 50 credits per day, each generation costs 10. So you can generate 5 times per day.
- Each time two versions are generated, each is about
1 minute2 minutes long. - There is Auto Mode and Custom Mode. Auto Mode is simple, Custom Mode is controllable.
- Lyrics should have at least two parts: Verse and Chorus.
- Chinese vocals can mispronounce some rare words and need to be replaced with simple words.
- After generating a song, you can continue to generate from it (Continue From This Song).
- After generating the second song, you can get the whole song (Get Whole Song).
I’ve tried it many times, but the first time I was most satisfied with the result. Although this Chinese vocal is a bit Cantonese flavored, it’s not bad at all. This was generated with the free v2, so I’m sure the paid v3 will be even better. Update @20240324: v3 is now available to all users!
About Lyrics
The lyrics is from a little poem I wrote few years ago, describing a true story of a 80s female friend. The whole poem is a bit long, and only the first three stanzas were chosen for this song demo.
The whole poem is here ( translated version ): Little Girl (Poem)
Chinese original version: 小姑娘 (诗)