Software Engineering Courses (SECourses)•3y ago

I tried one minute audio sample it didn't work then I tried 10 minutes samples that also didn't work

I tried one minute audio sample it didn't work then I tried 10 minutes samples that also didn't work,
in 1 minute audio trained data is empty , 10 minutes sample I can see only few lines in it

Ddns I tried one minute audio sample it didn't work then I tried 10 minutes samples t...

Furkan Gözükara SECourses•10/20/23, 3:46 PM

i used 2 hours training data

dnsOP•10/20/23, 3:47 PM

i only have less data files, is it possible to do from small files ?

Ddns i only have less data files, is it possible to do from small files ?

Furkan Gözükara SECourses•10/20/23, 3:59 PM

2 hours data is split into small files

Furkan Gözükara SECourses•10/20/23, 3:59 PM

how many small files you have

FFurkan Gözükara SECourses 2 hours data is split into small files

dnsOP•10/20/23, 4:02 PM

i have only one minute clip , and other file have 10 minute clip (this ten minute clip I merged multiple files into one using audacity)

Ddns i have only one minute clip , and other file have 10 minute clip (this ten minut...

Furkan Gözükara SECourses•10/20/23, 4:12 PM

i think that is why

Furkan Gözükara SECourses•10/20/23, 4:12 PM

probably you need minimum 10 training data

GeorgeClintonGeorgeBush•10/21/23, 6:23 PM

What's the current best way to generate yourself in images? SDXL or 1.5? Dreambooth or LORA?

GeorgeClintonGeorgeBush•10/21/23, 6:27 PM

Also, can a model of yourself have both your entire body and face accurately? Or does model face quality reduce if you include your whole body?

GGeorgeClintonGeorgeBush What's the current best way to generate yourself in images? SDXL or 1.5? Dreambo...

Furkan Gözükara SECourses•10/21/23, 9:07 PM

sdxl dreambooth is undisputedly best

GGeorgeClintonGeorgeBush Also, can a model of yourself have both your entire body and face accurately? Or...

Furkan Gözükara SECourses•10/21/23, 9:08 PM

we use adetailer and face details are still good

FFurkan Gözükara SECourses sdxl dreambooth is undisputedly best

GeorgeClintonGeorgeBush•10/21/23, 10:52 PM

Awesome, thanks for the info. I am new to SD. Is it possible to run SDXL Dreambooth on a 3060 12GB GPU or will I only be able to do a LORA?

GGeorgeClintonGeorgeBush Awesome, thanks for the info. I am new to SD. Is it possible to run SDXL Dreambo...

Furkan Gözükara SECourses•10/21/23, 11:01 PM

you can do sdxl drembooth on kaggle possibly. right now testing it

Furkan Gözükara SECourses•10/21/23, 11:01 PM

otherwise yes only lora

GeorgeClintonGeorgeBush•10/22/23, 4:13 AM

Okay, thanks!

GeorgeClintonGeorgeBush•10/22/23, 4:13 AM

I'll try LORA first

Nakedi Mafa•10/22/23, 6:29 PM

Hi @Furkan Gözükara SECourses I just came from your YouTube video on dolly 2.0 can you please help me with an error

NNakedi Mafa Hi <@205854764540362752> I just came from your YouTube video on dolly 2.0 can yo...

Furkan Gözükara SECourses•10/22/23, 6:44 PM

sure

Nakedi Mafa•10/22/23, 6:50 PM

ValueError: The current device_mapdevice_map had weights offloaded to the disk. Please provide an offload_folderoffload_folder for them. Alternatively, make sure you have safetensorssafetensors installed if the model you are using offers the weights in this format.

NNakedi Mafa ValueError: The current `device_map` had weights offloaded to the disk. Please p...

Furkan Gözükara SECourses•10/22/23, 6:51 PM

what is your GPU?

Nakedi Mafa•10/22/23, 6:53 PM

NNakedi Mafa Click to see attachment

Furkan Gözükara SECourses•10/22/23, 6:53 PM

ye probably you cant use with that

Furkan Gözükara SECourses•10/22/23, 6:53 PM

also try this web ui : https://github.com/oobabooga/text-generation-webui

GitHub

GitHub - oobabooga/text-generation-webui: A Gradio web UI for Large...

A Gradio web UI for Large Language Models. Supports transformers, GPTQ, AWQ, llama.cpp (GGUF), Llama models. - GitHub - oobabooga/text-generation-webui: A Gradio web UI for Large Language Models. S...

FFurkan Gözükara SECourses also try this web ui : https://github.com/oobabooga/text-generation-webui

Nakedi Mafa•10/22/23, 6:54 PM

Is this shown in the video from 8:55?

NNakedi Mafa Is this shown in the video from 8:55?

Furkan Gözükara SECourses•10/22/23, 6:58 PM

no this is not shown in video

FFurkan Gözükara SECourses no this is not shown in video

Nakedi Mafa•10/22/23, 7:01 PM

Do you have a video on how this works ?

NNakedi Mafa Do you have a video on how this works ?

Furkan Gözükara SECourses•10/22/23, 7:06 PM

not yet but will do hopefully

FFurkan Gözükara SECourses not yet but will do hopefully

Nakedi Mafa•10/22/23, 7:06 PM

Thank you very much for help

GeorgeClintonGeorgeBush•10/23/23, 3:39 AM

https://tenor.com/view/jay-skeleton-hmm-thinking-waiting-gif-17592139

Tenor

GeorgeClintonGeorgeBush•10/23/23, 3:41 AM

I'm using 45 training images and multiplying by 25 for regularization like in the tutorial. Too many training images?

GeorgeClintonGeorgeBush•10/23/23, 3:52 AM

Okay it's down to 29 hours estimated now

AndyWear•10/23/23, 5:20 AM

i discovered something pretty cool @Furkan Gözükara SECourses
While playing around with animateDiff, I was trying to trigger the camera words but did it wrong and instead of typing in the zoom in trigger word, I typed it in wrong and noticed the subject moving forward instead....
SO i thought, hey its prompting action words!!! So i prompted "running" and she ran.

Made me wonder how we can train action words?

AndyWear•10/23/23, 5:20 AM

Deleted User•10/23/23, 1:02 PM

Hello, anyone getting good results with voice cloning?

GGeorgeClintonGeorgeBush I'm using 45 training images and multiplying by 25 for regularization like in th...

Furkan Gözükara SECourses•10/23/23, 1:23 PM

depends on image quality

AAndyWear i discovered something pretty cool <@205854764540362752> While playing around w...

Furkan Gözükara SECourses•10/23/23, 1:24 PM

with action images and action captions

DDeleted User Hello, anyone getting good results with voice cloning?

Furkan Gözükara SECourses•10/23/23, 1:24 PM

i got this tutorial : https://youtu.be/OiMRlqcgDL0

YouTubeSECourses

Master Deep Voice Cloning in Minutes: Unleash Your Vocal Superpower...

Today, we're going to dive into the cutting-edge world of voice cloning and speech synthesis, a technology that has the potential to revolutionize communication as we know it. We're going to make use of TorToiSe TTS, an incredibly powerful yet user-friendly open-source tool for training and generating speech. And the best part? With TorToiSe TTS...

FFurkan Gözükara SECourses depends on image quality

GeorgeClintonGeorgeBush•10/23/23, 1:35 PM

Thanks. Estimate now is ending up at only 14 hours or so, down from 60.

Image quality is all good. I selected best from a set I took with a DSLR w/ 50mm

FFurkan Gözükara SECourses i got this tutorial : https://youtu.be/OiMRlqcgDL0

Deleted User•10/23/23, 3:23 PM

Thanks! Does it give good results? I tried few and was not able to get consistent results, I dont know how evenlabs do it

DDeleted User Thanks! Does it give good results? I tried few and was not able to get consisten...

Furkan Gözükara SECourses•10/23/23, 3:42 PM

eleven labs have better way

Furkan Gözükara SECourses•10/23/23, 3:42 PM

sadly not public