This library is built to synthesize ambient music based on emotion cues. These cues are passed to the system in the form of valence-arousal segments created from Russell's circumplex model.
The 5 segments shown below represent the 4 quadrants along with a "neutral" segment (#5).
The input to the system is a dictionary mapping of the various segments desired and the timestamps at which they are to be present (eg. {time1:segment1, time2:segment2, etc.}), and the system will synthesize an audio file matching this mapping, with smooth transitions from one segment to another.
-
Clone this repository using
git clone https://github.com/VedantKalbag/AmbientMusicSynthesis.git -
Open the folder and run
pip install -e .to make the folder importable (this will be simplified in a future release) -
The library can be used as shown below:
import ambientmusicsynthesis.ambientmusicsynthesis as ams
a = ams.AmbientMusicSynthesis()
a.generate_audio(90, {0.0:1, 20.3:2, 40.6:3, 60.9:4, 81.2:5, 90:1}, sample_rate=44100)
a.export_audio('test_n.wav')
Future work: The synthesis will have different themes for generation, which will control the overall type of the output. Examples include:
- minimal
- drone
- lush
- busy
- voice/choir style sounds