🎚️The Sliders (Prompt, Lyrics and Clarity)

The Big Three, and the subject of many an Udio debate!

This section will cover three core sliders within advanced controls: Prompt% Strength, Lyric% Strength and Clarity%.

Important to note before we continue some of the qualities of these sliders are not precisely known, and comments here are based upon analogical experience, not confirmation.

There are many different ways to use these sliders, and you'll have to find your own preferences. Some users will tell you there are "best settings" but this entirely depends on what kind of song you want, and you know that best! As with most Udio craft, experimenting for yourself is the best teacher!

Let's take a look at each slider in turn!

Prompt Strength

Always double check the value of this before generating, some options will reset it!

The Prompt Strength slider controls the prompt adherence the model will use. The higher the value, the more strictly you instruct the AI to stick to your prompt for the clip contents.

Think of this as the amount of "creativity" you allow the AI. At 100% it's essentially you telling the model "I want you to use my prompt for contents, and no added extras please!" At this setting the AI is being instructed to prioritise your prompt integrity, even at the expense of what sounds "natural or workable". This can cause it to "force" prompt elements together that normally won't go together, which can lead to interesting sounds, but can also sound artificial and unnatural.

At 0% you give the AI a lot of creative freedom with your prompt and how it chooses to interpret the terms. In such cases the AI will prioritise "what sounds workable" as opposed to obeying the prompt strictly. You're very unlikely to get strange or unnatural sounds at this setting, but the AI may totally ignore parts of your prompt.

Lyric Strength

This is a slider you really want to experiment with as it has some significant impacts on how vocalists perform.

The Lyrics Strength slider is another 0-100 slider which essentially determines how strictly the AI follows the lyrics in your lyric box. Higher values means it will read them with less deviation or creativity, lower values means it may read them more liberally.

It is incorrect to think of lower lyric values as meaning the AI will frequently ignore your lyrics. It may do this sometimes, but sometimes it may manifest as a slight slurred pronouciation or the like. The shorthand answer is the lyrical interpretation becomes less exacting which can make it sound more natural. This is important to consider when you're generating songs, fur in many genres this tendency for these "liberal flairs" within singing is part of the flavour.

Conversely higher lyrical strength % is very unlikely to have your lyrics ignored, but you can think of this as a result of the lyrics being force fed into the AI. This means accuracy will be high, but it could sound unnatural, particularly if the lyrical cadence isn't right, as the AI will force itself to perform them irrespective even if it sounds very clumsy. This can result in it bending words to fit them in.

This said if the lyrical cadence is good, this can result in some very interesting and very clean vocal deliveries that are excellent for certain genres. As keeps being said: experiment! Most people end up with a lyrical strength % they tend to prefer.

Clarity

Note: Clarity is not available when using the 1.0 model, so ignore this section if you're using it!

Clarity is again a 0-100 slider, with a default setting of 25%.

Clarity is the amount of separation of elements the generation will have. The higher the slider, the larger the separation. Larger separation will produce cleaner sharper sounds, but it could sound artificial and not suitable for certain genres. Lower clarity will reduce separation creating overlap of elements. This can make clips sound less crisp, but also more natural, particularly within certain genres.

Clarity tends to work best at different strengths for different genres, but it can also be used to make sections discreet from each other but having the "cleanness" of the clip change. Just bear in mind changing the clarity value significantly can result to changes in instrumental and vocalist sounds, so if you're not massively changing the structure of a section (or moving into a new one) significantly altering the clarity can alter how consistent a section sounds.

Experiment with this setting. All things being equal, altering clarity will produce different generations.