SpleeterAndroidPort

I spent multiple days to find a way to have an audio stem separation tool on Android, this repo contains a AudioStemSeparation class to use the converted tflite model of the 4stems spleeter model available in the jinay1991's repo from which this code was inspired (the tflite model can be downloaded here).

Configuration

• Add dependencies com.arthenica:ffmpeg-kit-audio, org.tensorflow:tensorflow-lite:2.13.0 and org.tensorflow:tensorflow-lite-select-tf-ops:2.13.0
• Place the tflite model in the assets folder of the app
• Place the AudioStemSeparation class where you need

Usage

val separator = AudioStemSeparation(context, "nameModelWithoutExt", "path/to/file/file.wav")
separator.init()
separator.separate()
separator.unInit()

If the file you are using is large and you get errors with memory consumption try adding android:largeHeap="true" to AndroidManifest under application tag.

The AudioStemSeparation class automatically saves the stems as 0.wav, 1.wav and so on.. representing bass, drums, accompaniment and vocals.