Like Dirt-Samples, but cleaned up, with clear provenance and license info (generally a permissive creative commons licence but check the metadata for specifics).
The bin/meta.py python script is a reference implementation that can make a '.cleanmeta' metadata file for your own sample pack folder. See below for how to use it and contribute a sample pack of your own.
If you want to use these outside the Tidal/SuperDirt/SuperCollider ecosystem you are very welcome. You're encouraged to join discussion in the github issue tracker so that we can develop a standard way to share and index/signpost these packs.
See /tidalcycles/sounds-repetition for an example sample pack which has two sets of samples in it.
Please only contribute samples if you are happy to share them under a permissive license such as CC0 or a similar creative commons license.
If you are unfamiliar with the 'git' software, please create an issue here, with a short description of your samples and a link to them and someone should be along to help shortly.
If you are familiar with git and running python scripts (or happy to learn), please follow the below instructions. This is all new - if anything is unclear please create an issue, thanks!
Create a '.cleanmeta' metadata file for each subfolder. Again, see tidalcycles/sounds-repetition for examples. There is a python script bin/meta.py which can generate the metadata file for you, run it without parameters for help. Here is an example commandline, that was used to generate repetition.cleanmeta:
../Clean-Samples/bin/meta.py --maintainer alex --email alex@slab.org --copyright "(c) 2021 Alex McLean" --license CC0 --provenance "Various dodgy speech synths" --shortname repetition --sample-subfolder repetition/ --write .
After generating the file, edit it with a text editor to fill in any missing info.
You can use free/open source software like audacity for editing samples.
As a minimum, be sure to trim any silence from beginning/end of the samples, and that the start and end of the sample is at zero to avoid clicks (you might need to fade in / fade out by a tiny amount to achieve this).
Consider adjusting the volume/loudness too, for example normalising to -1.0db - but this is very subjective and will depend on the nature of the samples and the music they're used with. For example distorted gabba samples are intended to be very loud, and a whisper is intended to sound silent. The average non-percussive sample should be around -23dB RMS. Samples shouldn't exceed 0dB true peak. EBU recommends -1dBTP at 4x-oversampling. Samples generally shouldn't have DC offset, although e.g. some kick drum samples naturally have non-zero mean.
For more advice, you could join the discussion here.
Thanks!