Out-performed HQ Stretching algorithm

Hello, I wrote an OpenCL-Kernel using app which is stretches audio faster than the high-quality stretching method in Audacity.

GitHub: MKLAudio github

Use the ‘timestretch_double03’ kernel with default chunkSize & overlap on an imported track.

Note: You’ll have to have OpenCL-Runtime installed!

GPUs currently do NOT work (use Intel / AMD CPU)

Audacity’s “HQ” time stretch is much slower than the “Standard” time stretch, but often sounds better and the processed length is precise. Being “faster” than the HQ time stretch does not automatically mean “better” unless the sound quality is at least as good and the stretched length is accurate.

Perhaps you could provide audio samples to demonstrate the quality with various kinds of audio.

This topic was automatically closed after 30 days. New replies are no longer allowed.