balthazarneveu / gyraudio

A novel method for audio source separation

1 stars 0 forks source link

Generalization : infer any signal #21

Closed balthazarneveu closed 6 months ago

balthazarneveu commented 6 months ago

Infer any signal

Combine any noise and any signal (switch signal using 8/2 , switch noise using 9/3)
Resample to the expected samplin rate 8kHz.

Example:

python scripts/audio_separation_interactive_custom.py 
--preload
-e 1002
--noise "__data_source_separation/source_separation/test/00*/noise.wav"
--signal "__data_source_separation/real_voices/crocolou_8kHz.wav" "__data_source_separation/real_voices/mon_code_secret_8khz.wav" "__data_source_separation/voice_origin/test/Jules_Verne_-_L_ile_mysterieuse_Partie1_Chap15_40_.wav"

att-hack compare male vs female voices.

python3 scripts/audio_separation_interactive_custom.py -n "__data_source_separation/source_separation/test/00*/noise.wav" -s   __data_source_separation/att_hack/Volumes/CLEM_HDD/IRCAM/Open_SLR/wav/M17_a4_s006_*.wav __data_source_separation/att_hack/Volumes/CLEM_HDD/IRCAM/Open_SLR/wav/F01_a4_s006_*.wav  -e 1002 --gui qt -p

background sounds audio datasets in general

Interactive version

[x] infer any voice in the GUI
[x] add info title
[x] resampling

Batch processing

-> infer any voice in the batch - moved to #23