Iver Jordal

icon company ElevenLabs icon location Trondheim, Norway icon location Machine learning, audio/music tech, computer vision, demoscene, web technology, games, startup. I mainly write Python. I also enjoy Go, C, Rust and JS.

Results 255 comments of Iver Jordal

Add gain peak

Thanks for the patience. This is still on my TODO list 🙈

Add gain peak

I had another look at this now, and gathered some thoughts. Here's what I imagine would be good to have in/for this transform: - [ ] The peak (offset) can...

Vocal tract length perturbation

Closing for inactivity. Feel free to suggest a reopen later if the work gets picked up again

Implement compressor

This one has appeared: https://github.com/yoyololicon/torchcomp I haven't tried it yet, but it looks promising at first glance.

add randomized shift

Could you write a few sentences that explains what this transform does, what it sounds like and when it is useful?

Avoid picking silent part of noise file

Thanks for sharing! If I were you, I would "clean" the data, i.e. remove the silent part from the files, and then that would be the fix for the observed...

Avoid picking silent part of noise file

Good point. AddShortNoises may indeed be better suited for your need. You could set [noise_transform](https://iver56.github.io/audiomentations/waveform_transforms/add_short_noises/#noise_transform) to `Trim()`, and it would trim your noise sound before finding an offset for it...

Add support for models with lower precision (like 16-bit floats and 8-bit integers)

Also try precision=16 in pytorch-lightning. By default it uses the native amp in pytorch

Process hangs, out of memory?

One of the reasons why pyloudnorm uses a lot of memory is that it creates a float64 copy of the audio internally for filtering, due to the way scipy works....

Change default noise_rms="relative_to_whole_input" in AddShortNoises

I can close this issue after we make the final switch. For now it just says this: ``` warnings.warn( "The default value of noise_rms in AddShortNoises will change from" "...