# talknet-support
- w
WeegeeFan1
10/29/2022, 7:10 PMMaybe I don't have any memory left - w
WeegeeFan1
10/29/2022, 7:17 PMyeah i cant figure it out ill wait for someone to say something - w
WeegeeFan1
10/29/2022, 7:25 PMI'll try one more time making the save spot the root directory of my drive - w
WeegeeFan1
10/29/2022, 7:37 PMnow im restarting again because putting it as the root folder is baaaad.... I'm deleting everything in my drive and moving it to my harddrive to see if thats the issue and ill retrain one more time - w
WeegeeFan1
10/29/2022, 7:38 PMOtherwise I'm giving up - w
WeegeeFan1
10/29/2022, 9:27 PMYES IT WORKED I DIDS IT - w
WeegeeFan1
10/29/2022, 9:27 PMORJNKSLKMNTFKB - w
WeegeeFan1
10/29/2022, 9:27 PMOk now I need to test it - w
WeegeeFan1
10/29/2022, 9:46 PMSo I managed to use up all of the GPUs I can use for free with Collab. Is there a way to manually mount it to my own GPU? Would an RTX 2080 work? - w
WeegeeFan1
10/30/2022, 3:50 AMI made my first model. It has seizures as I expected - w
WeegeeFan1
10/30/2022, 3:50 AMI made a mire complete model with a good 50 minutes worth after - w
WeegeeFan1
10/30/2022, 3:51 AMBut I don't think I have the space on my drive fir the notebook to train it - w
WeegeeFan1
10/30/2022, 3:51 AMShould I make the notebook saved in the drive? IS that bad? - s
Skyler
10/30/2022, 6:24 PMabout how many minutes of singing can make a good talking model into a good singing model via multispeaker - b
Big Lez
10/30/2022, 7:25 PMYo, just trained up a voice... is there a way I can then get that voice to do reference audio? Or am I limited to the reference audio voices on Controllable TalkNet? - w
WeegeeFan1
10/30/2022, 10:19 PMYou can add to the files of the collab then load them with custom reference audio. - w
WeegeeFan1
10/30/2022, 10:19 PMJust go on yt search how to use the controllable talknet - w
WeegeeFan1
10/30/2022, 10:20 PMI like the controllable takknet. It let's you use the features of uberduck talknet without the limits of uberduck - l
Lexi (delulu posts on the daily)
10/30/2022, 10:20 PMon controllable talknet look for the dropdown it'll say custom model then select that and put in the google drive id - w
WeegeeFan1
10/30/2022, 10:21 PMPrecisely - b
Big Lez
10/30/2022, 10:27 PMNice one thanks! Looking forward to getting it done tomorrow. 😊 - w
WeegeeFan1
10/30/2022, 10:28 PMMy dad let me use his Google drive where he has more storage so I can train my models. - w
WeegeeFan1
10/30/2022, 10:28 PMI don't have issues anymore - w
WeegeeFan1
10/30/2022, 10:29 PMYes I'm younger but very tech savvy - w
WeegeeFan1
10/31/2022, 12:13 AMI want to know if it's possible to train it locally on your own hardware. That would be so usefull.. - j
Justin
10/31/2022, 2:38 AMIt is possible - j
Justin
10/31/2022, 2:38 AMBut not easy to setup - j
Justin
10/31/2022, 2:38 AMEspecially on Windows - b
Big Lez
10/31/2022, 11:19 AMJust got this error for reference audio with custom trained model, any ideas? - w
WeegeeFan1
10/31/2022, 8:27 PMI'm on Windows 10 and have an RTX 2080. How would I set that up?