A demonstration of using AI to generate custom audio. Both examples are readings of the phrase in the box below.
The first example comprises two AI generated words, name and colour, interspliced with two pre-recorded sentence fragments. The second example is entirely generated by AI.
I think the second example shows that when we want to customise names, etc, it would be preferable to generate the entire phrase using AI instead of replacing the odd word. The app would then intersplice these generated phrases with pre-recorded audio. I think the second example sounds much more fluent than the first.
Note: this page is slow to load because it needs to wait for the AI speech to be generated. In production, AI generated phrases would be cached and eager-loaded to keep load times low.
Once there was a dog named Gordon, and its collar was a lovely shade of green.
The audio below is the above phrase with only the name and colour generated by AI.
The audio below is entirely generated by AI.
Customise the audio examples with a new name and colour.