Can I… like… pay someone to make a LORA of me? I’m running out of steam/patience
Hey guys,
I’ve just spent the last 2 days (albeit part time) trying to train a LORA model of myself either through a local Dreambooth or on Google Colab through [Kohya_ss-colab](https://github.com/camenduru/kohya_ss-colab) while following [this great tutorial](https://www.youtube.com/watch?v=70H03cv57-o) and all I’m gonna say, long story short, is if I’m not running into errors, the safetensors I get simply don’t work once I test them out (i.e. it’s deep-friend or smudged non-sense, even after reducing training steps).
My dataset of 20 image and caption files are more than decent (as far as I’m aware). Yet, clearly, I must be doing something wrong, because nothing ever seems to work as anticipated. I’ve trained many models in the past with no issues, but LORAs don’t cooperate with me for some reason.
Since I’ve run out of patience, I thought I’d just try and see if anyone with an already-working LORA training setup would be interested in training a LORA of me for a fee (PayPal only)?
I would provide you with a ZIP file containing the 20 images + 20 caption files (all properly named) and all I’d ask in return is the .safetensors file back.
If interested, please shoot me a DM with your fee and estimated deadline.
Thank you very much! Much appreciated.
[This guy takes commissions.](https://civitai.com/user/YeiYeiArt/models)
i made a post last week about how i did mine, theres a link to another tutorial which works. i didnt even bother to caption my 23 images and the results are pretty consistent. took 4h on a 3090 but im pretty sure i could have stopped half way.
If you have the dataset, I can do it for free. I’m currently testing different settings for fine tuning as well as the ability for my checkpoint to be easily trained. I’ll just extract a lora from it and give it to you. I don’t guaranty the result on other checkpoint.
Just faceswap with roop?
I’ve had some success with [a couple of loras](https://civitai.com/user/demoran/models).
I use wd14 captioning, and prefix the caption files with the trigger word (eg YourName).
I’ve been running 10 epochs for about 7000 steps. I’m usually working with more than 20 images.
After the lora is done baking, I put it through its paces, first on a checkpoint at different weights, and then through multiple checkpoints with wildcards to get representative images and see if anything wonky is going on.
My machine takes about an hour and a half to finish creating a lora.
Honestly, most of the effort has been finding good pictures in a sea of dross.
One strategy I haven’t really tried yet is taking a video and getting stills from it and making a lora from that. I got this program that will grab stills every n frames, but again “sea of dross”