Really insightful, apologies for being crude and shallow but (hindi speaker) it was hard to focus without giggling at "Train Your Lora" so many times😂😂😂😂
Oh love this! Needed something like this for thumbnails lol. Dont want to click a new photo everyday for a thumbnail haha! BTW the video background is coming from photoshop GEN FILL or some other app ? Looks quite nice and clean! Did you make a video about this same ?
Really cool and simple method. Thank you for sharing 'boson'! BTW I see you are using Mac, so is it possible to have a stable AI open-source setup on a Mac like SDXL etc.? Thank you.
Yes, you can. That’s why it’s a good practice to add your hugging face repo details and secret key. You will find the safetensors file on HF which you can use locally.
Can you create a similar tutorial to train the lora locally on A1111U, i have flux or sd models installed locally, how do i train and add my lora there without using replicate
The code itself is free for non-commercial use. You're paying for the GPU time for training and inferencing. There are local version you can run if you have the requisite GPU
When I click "Create training" I get this message - "You need to have a payment method set up to run this model" however I can't find where to set this up. Can you please advise
If I understand this correctly, can I use this safetensors file that you created in the end in A1111 as well? Also can this training be done on a local machine instead of replicate?
I am getting this error : Prediction failed. 401 Client Error. (Request ID: Root=1-67408e2b-4df9ef7237663b6e0d7137bf;5c6df82d-550a-4978-9a4c-89dfa3c5ac4d) Repository Not Found for url: Please make sure you specified the correct `repo_id` and `repo_type`. If you are trying to access a private or gated repo, make sure you are authenticated. Invalid username or password. tried to solve whats it saying but still not
Imagine this model being used to create a deep fake of any person.( without their consent) It could have significant impact. How do you deal with these kind of issues?
There are laws that keep those issues in check - no AI specific law has come out. However, it can fall under identity theft, misappropriation and other cybercrime related laws. Took a while for cybercrime laws to emerge, same goes for AI regulations as well.
Your input images were probably bad or you may have missed a step. Also it takes 30 mins max, not sure why it took 2 hours. We’ve trained at least 5 models like good and we continue to get good results.
@@100xEngineers I am pretty sure I didn’t miss any step. I was thorough I took my time to find good 60 images of mine with different angles and lighting thats why it took longer than 30 mins. The model took around 30 mins just to train. Can you share what dataset you used? I can try to replicate similar face angles. I might have overdone the dataset. Maybe I used too many angles.
Yo man, it's not even worked at all. I tried couple of times and they charged me 5 to 8 dollars on each attempt. Idk either you missed any steps or this process is fucked up for sure. I have tried with other models but not in this way, but this is not working not even for a single time.
Hey 100x , awesome video! Your unique style really stands out. I'm a video editing specialist focused on TH-cam growth, and I can help enhance your content with high-quality edits and engaging thumbnails. If you're ever open to exploring new ideas or strategies that can boost your channel further, I'd love to connect and see how we can collaborate. Let me know your thoughts
7:01 I'm definitely using this line with the homies
"so we're actually done training on your face??"
@grzwl 🤫
Never been more excited to use my lora
😂
😂😂😂
exactly, sister.
😂
9:14 no way you said that with a straight face 😭
Really insightful, apologies for being crude and shallow but (hindi speaker) it was hard to focus without giggling at "Train Your Lora" so many times😂😂😂😂
This is just the beginning 🔥
Best TH-cam channel ever!
Well explained 🔥🔥
Oh love this! Needed something like this for thumbnails lol. Dont want to click a new photo everyday for a thumbnail haha!
BTW the video background is coming from photoshop GEN FILL or some other app ?
Looks quite nice and clean! Did you make a video about this same ?
how much are you willing to pay for a service like this that does better quality?
Do you need to pay first before you can run the Training on replicate? thanks
enter prompt: smart successful young indian youtuber sitting at a desk with laptop curly tall hair, his grin remains consistently enigmatic
Yeah FLux is dope open source model you can run it locally as well, will require bit of ram and will be slightly slow that's it
How much RAM and GPU is enough?
16GB VRAM minimum is required for flux dev model
Hmmm, deleted in less than 5mins, okay. I see where 100X is drawing the line. Fair. I m with you guys. ❤
deleted what?
It was about voice cloning with few simple steps but got deleted
Really cool and simple method. Thank you for sharing 'boson'!
BTW I see you are using Mac, so is it possible to have a stable AI open-source setup on a Mac like SDXL etc.? Thank you.
It’s possible - not the most ideal.
This is a gems.. Could you please create a video about Voice.
he did but he deleted
compare with Shakker ai ? i confuse to use your way and shakker .
6:30 wow this is so affordable, I have doubt, how much will it cost to generate 1000 images ? is there any limit ?
Solid video, Thank you mate.
What was your dataset size for training Amul campaign thing? Does it detects the texts while training and generate ?
About 15ish images. Yes, it screws up the text. But gets the style right
Ok bro nice vedio so how much money we need to spent on this to get this one ❤
rather make a checkpoint than a lora for this tbh, better quality
so was it not free ?
first i faced problem in uplading file
Second payment problem
This is bomb. Thanks for sharing ❤️
Wow super bro make these kind of vedios more
Can i download this Lora that i've trained and use it in local stable diffusion?
Yes, you can. That’s why it’s a good practice to add your hugging face repo details and secret key. You will find the safetensors file on HF which you can use locally.
@@100xEngineers Thank you
Why don't you upload frequently, great video.
Do you have some other channel as well?
Can you create a similar tutorial to train the lora locally on A1111U, i have flux or sd models installed locally, how do i train and add my lora there without using replicate
What prompt did you use to generate an image of yours in the thumbnail ?
I am getting an error "Failed to training", can someone help me resolving this?
You need to add credit card details first. Then this error won't come
Can you show using colab!
I can't able to upload the file 😢
is it free?
today i was trying to create model and upload zip folder, nothing happening, many times i tried but not succsess.
subscribed❤❤
once model is created and trained, can we finetune or improved it?
It’s already a fine tuned model.
He means finetuning it further to include other people faces
@ yes correct
How can something be open-source if it requires a fee for accessing the code?
The repo is opensource on GitHub. The fee is for running the model on Replicate GPUs. If you have your own local GPU, then it’s free of cost for you.
The code itself is free for non-commercial use. You're paying for the GPU time for training and inferencing. There are local version you can run if you have the requisite GPU
🤦🏾
When I click "Create training" I get this message - "You need to have a payment method set up to run this model" however I can't find where to set this up. Can you please advise
Go to Account Settings, Billing
@@100xEngineers Thank you. I got it sorted
Is this free to use?
If I understand this correctly, can I use this safetensors file that you created in the end in A1111 as well? Also can this training be done on a local machine instead of replicate?
Yes, you are correct. It can be done on a local machine if you download the repo and run it, yes.
Wonderful ..
Let me know once you able to do it on local.
I am getting this error : Prediction failed.
401 Client Error. (Request ID: Root=1-67408e2b-4df9ef7237663b6e0d7137bf;5c6df82d-550a-4978-9a4c-89dfa3c5ac4d) Repository Not Found for url: Please make sure you specified the correct `repo_id` and `repo_type`. If you are trying to access a private or gated repo, make sure you are authenticated. Invalid username or password.
tried to solve whats it saying but still not
good stuff.
Appreciate the support!
Any way to train for free online?
you can also create all this for free but you gonna need to run locally in your system
How ?
how to get the background like you, its not real right, then how to make one
@@user-qx9vw5tm8h ya bro but i want the exact background can you give me the prompt
How can you say this model's name with a straight face ? 😂
Add timestamps bro
Yeah bet ill just use it on my own face. (proceeds to use make lora out of henry caville)
❤❤❤
🎉
Anyone who has tried Flux Lora to replicate products instead of faces? How does it perform?
LoRA stands for…… 😂
I am definitely going to hell. 😂
Boys , Let's learn how to use L O R A
Imagine this model being used to create a deep fake of any person.( without their consent) It could have significant impact. How do you deal with these kind of issues?
There are laws that keep those issues in check - no AI specific law has come out. However, it can fall under identity theft, misappropriation and other cybercrime related laws. Took a while for cybercrime laws to emerge, same goes for AI regulations as well.
I got extremely bad results. 2 hours which I am not getting back.
Your input images were probably bad or you may have missed a step. Also it takes 30 mins max, not sure why it took 2 hours. We’ve trained at least 5 models like good and we continue to get good results.
@@100xEngineers I am pretty sure I didn’t miss any step. I was thorough I took my time to find good 60 images of mine with different angles and lighting thats why it took longer than 30 mins. The model took around 30 mins just to train.
Can you share what dataset you used? I can try to replicate similar face angles. I might have overdone the dataset. Maybe I used too many angles.
Yo man, it's not even worked at all. I tried couple of times and they charged me 5 to 8 dollars on each attempt. Idk either you missed any steps or this process is fucked up for sure. I have tried with other models but not in this way, but this is not working not even for a single time.
That’s really weird. I’ve trained 4 models using the same method over here. Maybe try distant epoch cycles - 1500 perhaps.
Correct its not working... Always getting prediction failed. Error 401.....
LORA 🤣
6:29, this is really *cheap*
7:01 💀
🗿
Would you have a $10 free invitation to replicate, please?
is this paid
It still looks fake :)
old content
Bruh this lora name needs to be changed
It’s not free guys 😅
Hey 100x , awesome video! Your unique style really stands out. I'm a video editing specialist focused on TH-cam growth, and I can help enhance your content with high-quality edits and engaging thumbnails. If you're ever open to exploring new ideas or strategies that can boost your channel further, I'd love to connect and see how we can collaborate. Let me know your thoughts
Email on hello@100xengineers.com with your portfolio.