In the current implementation, we chunk the enhanced segment into the original cut boundaries and save it to output files in the same loop as the batch processing. This probably makes the GPU wait while the post-processing is finished. We can probably submit the post-processing to an async function and continue with the enhancement loop to provide a little speed-up.
In the current implementation, we chunk the enhanced segment into the original cut boundaries and save it to output files in the same loop as the batch processing. This probably makes the GPU wait while the post-processing is finished. We can probably submit the post-processing to an async function and continue with the enhancement loop to provide a little speed-up.