Now you can describe a sound and this model will generate audio samples corresponding to your text. Total four samples are generated. Add more details in description to get better results.
Lets check out some examples:
First let us use a straight forward prompt for our audio sample without adding any details say - Walking in mud.
As you can see among four samples that were generated, above one was closest to walking and the audio sounds more like walking on road. We want sound of walking in wet, muddy region like in a forest after heavy rain. So, lets add some adjectives to describe this sound better.
Squishing or squelching noise of walking in wet mud
Now above one came out accurate. So let us try another example. This time we thought of generating horror sounds, and we added some adjectives to describe it as below.
Horror whispers consist of soft, hushed sounds, such as breathing, muttering, murmuring, voice like or more abstract