I like to extend from my own demos to make it more organic, then i cover the ones i like for a complete song. But the lyrics....a lot of shadows, whispers, neon, storms, electric, glow, sparks, code, echoes, etc. Same with deepseek.
My dinosaur band lyrics is often better. The prompt is usually something like "a childrens rock song about a t-rex eating too much pizza".
I’ve found a Chat GPT trained on theology proper when given more direction on lyrics from a specific bible passage creates much better lyrics than just asking Suno to create them. Then giving those lyrics to Suno makes a superior product. It’s all about the inputs.
The audio sounds compressed and two correct. There are no errors. It is the difference between a master potter who never quite gets things perfect and the 3d printer replicating things precisely.
I think we are going to end up with a lot of slop writing and music because the creator does not take the last step of hand finishing the framework and recreating/revising the work until it fits within their style.
AI will make a lot of things, true. But if you do not have the organic or acoustic skills to remake things, it will not reach high art.
The output vision improves as personalization and the AI's cross-exchange 'memory' of the user grows. Even within a single exchange, you can share your vision before asking for lyrics and musical direction to cross-pollinate in SUNO. (Make sure to tell it it's for use in SUNO.)
Also, an extremely short request with the auto lyrics settings in SUNO produces lyrics that sound like a cross between ipsum lorem and a stroke. I hope an example link is acceptable. https://suno.com/s/7w63OZocwJEmxsga
Yeah, I was trying to write a cyberpunk racing song, there was the word "neon" on every other line. There are two features I'd welcome, I don't know if they are present in the paid version.
1. More control in the music generation. I want to be able to specify scale and chord progression.
2. I want to be able to edit the lyrics and potentially the music on the fly like Claude does with text. Right now if I try to slightly edit a result I liked, I get a completely different song, that's potentially worse.
I also found a guy on youtube who samples the voices of actors and singers and is able to incorporate them with Suno. I haven't looked deeper into it, but from what I saw it wasn't very complicated, so I'm guessing the next copyright lawsuit is coming from singers and voice actors.
It's possible to use the editor and edit certain sections as a rough cut, even rearranging, then export the song. Then you can reimport the song into Suno for a more polished product using it as a style to copy by the AI.
I just asked chat gpt to write me a song like apt and the first line is
“We’re dancing in the shadows of a neon dream” 😂
I like to extend from my own demos to make it more organic, then i cover the ones i like for a complete song. But the lyrics....a lot of shadows, whispers, neon, storms, electric, glow, sparks, code, echoes, etc. Same with deepseek.
My dinosaur band lyrics is often better. The prompt is usually something like "a childrens rock song about a t-rex eating too much pizza".
I’ve found a Chat GPT trained on theology proper when given more direction on lyrics from a specific bible passage creates much better lyrics than just asking Suno to create them. Then giving those lyrics to Suno makes a superior product. It’s all about the inputs.
This also applies to written works. Adding a short story with the prompt yielded a better output story than solely the prompt.
The audio sounds compressed and two correct. There are no errors. It is the difference between a master potter who never quite gets things perfect and the 3d printer replicating things precisely.
I think we are going to end up with a lot of slop writing and music because the creator does not take the last step of hand finishing the framework and recreating/revising the work until it fits within their style.
AI will make a lot of things, true. But if you do not have the organic or acoustic skills to remake things, it will not reach high art.
The output vision improves as personalization and the AI's cross-exchange 'memory' of the user grows. Even within a single exchange, you can share your vision before asking for lyrics and musical direction to cross-pollinate in SUNO. (Make sure to tell it it's for use in SUNO.)
Also, an extremely short request with the auto lyrics settings in SUNO produces lyrics that sound like a cross between ipsum lorem and a stroke. I hope an example link is acceptable. https://suno.com/s/7w63OZocwJEmxsga
That's an apt description. It could also apply to various arguments I've heard people make.
On a more meta level, the use of AIs specialized for specific tasks, and then "cross pollinating" is an effective and intermediate technique.
If you want to take it to the next level, you can start automating processes, chaining multiple AIs into one workflow.
n8n Cloud is a relatively straightforward way to do this and you can download some templates along with seeing many possibilities here:
https://www.youtube.com/@nateherk/videos
Lots of SUNO users use the cross-pollinating technique. They even leave the icons in the entered lyrics.
Yeah, I was trying to write a cyberpunk racing song, there was the word "neon" on every other line. There are two features I'd welcome, I don't know if they are present in the paid version.
1. More control in the music generation. I want to be able to specify scale and chord progression.
2. I want to be able to edit the lyrics and potentially the music on the fly like Claude does with text. Right now if I try to slightly edit a result I liked, I get a completely different song, that's potentially worse.
I also found a guy on youtube who samples the voices of actors and singers and is able to incorporate them with Suno. I haven't looked deeper into it, but from what I saw it wasn't very complicated, so I'm guessing the next copyright lawsuit is coming from singers and voice actors.
It's possible to use the editor and edit certain sections as a rough cut, even rearranging, then export the song. Then you can reimport the song into Suno for a more polished product using it as a style to copy by the AI.
1. scales and chord progressions are more miss than hit.
2. You can replace sections, but again, it's hit or miss. Nowhere nearly as easy or seamless as Claude.