Authors

Higgue
Dataset Extraction, Mod Compiler, Additional Vocal Coaching

Allen Scott
Input Vocal Performances, Audio FX (Rebecca's Megaphone FX)

DarkShift
Audio FX Assistance (Shade's Radio FX)

AI Training Dataset Source

PAYDAY 3
Intellectual Property

Starbreeze Studios
Developing Studio

Overkill Software
Developing Studio

Deep Silver
Publisher

Character Voice Actors

Simon Kerr
Dallas

Damion Poitier
Chains

Pete Gold
Hoxton

Nicklas Berglund
Wolf

Derek Ray
Houston

Carmen Duran Garcia
Damien Faity Rougier
Antonella Giannini
Holger Buhr
Jacket

*Pulled from the wiki, official credits for Jacket not directly listed. The female operator voice is the only one actually sampled and used in our mod, which is Carmen Duran Garcia as far as we know.

Siu-see Hung
Joy

Rebecca LaChance
Pearl

Aoife Duffin
Clover

Kosha Engler
Shade

Jill Winternitz
Rebecca (FBI Van)

Derek Hagen
Henry Garett
Mike Bodie
Jason Keller
Cops 1 - 4

Software

Mangio-RVC-Fork
AI Training

MMVCServerSIO
Realtime AI Voice Recording

Dallas
Model inference was done in realtime, feature index was not utilized.
1950 Epochs worth of training.
Dataset was 7 minutes and 46 seconds of speech. (All Assault VO)

Chains
Model inference was done in realtime, feature index was not utilized.
800 Epochs worth of training.
Dataset was 4 minutes and 32 seconds of speech. (All Assault VO)

Hoxton
Model inference was done in realtime, feature index was not utilized.
850 Epochs worth of training.
Dataset was 5 minutes and 27 seconds of speech. (All Assault VO)

Wolf
Model inference was done in realtime, feature index was not utilized.
2600 Epochs worth of training.
Dataset was 9 minutes and 19 seconds of speech. (All Assault VO)

Houston
Model inference was done in realtime mostly, one or two were done through Mangio-RVC-Fork. The feature index was utilized for those few.
1560 Epochs worth of training.
Dataset was 9 minutes and 54 seconds of speech. (All Assault VO)

Jacket
Model inference was done through Mangio-RVC-Fork. The feature index was utilized at a ratio of 1.
3045 Epochs worth of training.
Dataset was 4 minutes and 5 seconds of speech. (All Female Operator VO)

Joy
Model inference was done in realtime, feature index was not utilized.
1125 Epochs worth of training.
Dataset was 4 minutes and 45 seconds of speech. (All Assault VO)

Pearl
Model inference was done in realtime partially, the rest were processed in post using Mangio-RVC-Fork. The feature index was utilized for most of the lines.
1600 Epochs worth of training.
Dataset was 4 minutes and 2 seconds of speech. (All Assault VO)

Clover
Model inference was done in post using Mangio-RVC-Fork. Feature index was utilized at a ratio of 1.
1800 Epochs worth of training.
Dataset was 10 minutes and 6 seconds of speech. (All Assault VO)

Shade
Model inference was done in realtime, feature index was not utilized.
1625 Epochs worth of training.
Dataset was 7 minutes and 45 seconds of speech.
Radio filter was added in post by DarkShift using Ableton.

Rebecca (FBI Van)
Model inference was done in realtime, feature index was not utilized.
625 Epochs worth of training.
Dataset was 5 minutes and 20 seconds of speech.
Prior to training, Mangio-RVC-Fork was used to DeReverb the speech.
Reverb and Echo reapplied in post by Allen Scott using Audacity.

Cop 1
Model inference was done in realtime, feature index was not utilized.
500 Epochs worth of training.
Dataset was 7 minutes and 41 seconds of speech.

Cop 2
Model inference was done in realtime, feature index was not utilized.
1225 Epochs worth of training.
Dataset was 7 minutes and 52 seconds of speech.

Cop 3
Model inference was done in realtime, feature index was not utilized.
1050 Epochs worth of training.
Dataset was 7 minutes and 19 seconds of speech.

Cop 4
Model inference was done in realtime, feature index was not utilized.
625 Epochs worth of training.
Dataset was 11 minutes 11 seconds of speech.