Riffusion finetuned with google/musicCaps.</br> I found that prompt similar to dataset label returns more fidable result.</br> In my case I made a prompt with Chat GPT like this: <p> I'm writing prompt for music generation ai. I used captions like this:
- someone is playing a high pitched melody on a steel drum. The file is of poor audio-quality.
- This is a glitch music piece. There is a synth sound rising in pitch that resembles a triangle wave. There are granular synth samples being played randomly. A virtual percussive low-to-mid bell sound is playing a melody that resembles a marimba. There is an eerie feeling of flow. This piece could be used in the soundtracks of dystopian sci-fi movies. It could also be used in exploration sequences of video games.
- This file contains an orchestral composition rising up while a lot of digital clicking sounds are in the foreground. This is an amateur recording. And the sounds seem to come from a different source. This song may be playing in an adventure videogame.
Now I want to make a soothing jazz with base with medium temp. Write a propmt in styles similar to above captions. Return one sentence with 3 lines </p>
Response: Create a serene dance atmosphere with a dreamy melody, soothing synths, and a pulsing beat that gently propels listeners into a state of blissful tranquility, perfect for unwinding after a long day or enjoying a moment of peaceful dance.
