🗒️Core Prompt Box
The Core Prompt Box is where a lot of the main magic happens!
The Core Prompt Box (CPB) is where you tell the Udio model what you want your song to be like. It's the space where you can express things like genre, instruments, vocalist types and mood.
There are many different ways to use the CPB. Here we'll explain a few different ways it can be used. You should experiment with a few different methods and find what works best for you! Before that though, there are few things that should be explained!
Firstly in the top right corner of the CPB is a small dice icon: 
This is the randomize prompt tool. If you click this the CPB will automatically be filled with a prompt generated by the Udio AI itself:
Use this tool to help get you started and also see some of the different ways you can prompt.
The second major feature to be aware of are recommended tags. As you type, the Udio model will automatically recommend tags for you which harmonise/align with what you seem to be typing. They will appear underneath the CPB typing area. You don't need to click them, but if you do they will automatically be added into the text of the CPB:
Now let's talk about different ways to prompt! We'll keep things relatively simple here but the best advice is to experiment. You'll be surprised what the model can work with!
Simple Phrases
This method would be prompting along the same lines as the recommended tags work, single phrases separated by commas, like this:
Pop, Female Vocalist, Energetic, Electronic, Uplifting
The recommended tags will give you an idea of what kinds of phrases can be used, but generally speaking it's things like genre, vocalist(s), moods, instruments and structural elements. You can type things here the recommended tags will not recommend, and the model may still work with them - you have to experiment!
Take a look at the Knowledgebase for a collection of different phrase tags you can use in the CPB!
Descriptive
This method would be writing in sentences in a way that is more like given instructions, as if you're talking to someone and telling them "what to do":
An upbeat and energetic pop song with a female vocalist, electronic beats with a danceable rhythm, the song starts with minimal instrumentation and moves into heavier electronic beats as it progresses into the chorus...
This method of prompting is very open-ended. It works completely outside of the same system as the recommended tags so there is less certainty the model knows "what you mean" when you prompt like this, but there is definitely evidence it does respond to descriptive prompts. You may need to think carefully about how clear your wording is!
You can mix and match these prompt styles as follows:
An energetic and upbeat pop song with a very danceable and catchy chorus, female vocalist, electronic, melodic, choruses progress into heavy electronic drops...
Be creative! As the image regarding the randomize prompt showed above, you can type very abstract things into the model and see what happens!
As has been said many times, experiment, find what works for you and what gives you results along the lines of what you like! Udio is very much a tool the more you tinker, the more you're going to get out of it! The more you practice with different prompting styles and terms, the more you're going to understand how to control what kinds of results you get!
Last updated