Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

What Videos Do You Want To See Next? (Patreon)

Published:
2023-10-30 16:06:14
Imported:
2026-06
Tags:

Poll

What Videos Do You Want To See Next?

  • Dreambooth Training 226
  • Other LoRA Training (tell me what in the comments) 98
  • Text-Generation Webui Install Videos 91
  • Complete Stable Diffusion Usage For Beginners 163
  • Others (Let me know in the comments) 37
  • 2023-10-30
  • 615 votes
{'title': 'What Videos Do You Want To See Next?', 'choices': [{'text': 'Dreambooth Training', 'votes': 226}, {'text': 'Other LoRA Training (tell me what in the comments)', 'votes': 98}, {'text': 'Text-Generation Webui Install Videos', 'votes': 91}, {'text': 'Complete Stable Diffusion Usage For Beginners', 'votes': 163}, {'text': 'Others (Let me know in the comments)', 'votes': 37}], 'closes_at': None, 'created_at': datetime.datetime(2023, 10, 30, 13, 38, 36, tzinfo=datetime.timezone.utc), 'description': None, 'allow_multiple': False, 'total_votes': 615}

Comments

Retired Android

Training concept lora. For example, training lora that features holding of something, fantasy medieval weapon or such.

Harvey Webb-Jones

Openpose or controlnet posing for SDXL. Would be a massive help towards bulkier models like space marines

Gluthoric

Control Net Guide, Some of them I just don't understand XD I mean I get the concept but would like to just sanity check myself

Jonathan Streeter

Guide to adaptive optimizers in Kohya Lora training in 1.5 - Prodigy is supposed to be the best but I can’t find any real guides.

Aitrepreneur

concept is very similar to style training, just need a bunch of very different images that represent the concept well, then use the style preset training

Chrispy

I really want to see a ComfyUI workflow tutorial. ComfyUI is more powerful than Automatic1111, and achieves MUCH better speed on users of Nvidia GPUs with the latest drivers. But, it is significantly more complex once you get past the surface, and could use some explanation for things like adding Loras, adding ControlNet, or performing inpainting/outpainting. Otherwise, a series of videos explaining the most effective uses for the different ControlNet models would be fantastic.

ZenzeosArt

image-Tagging with ChatGPT 4V or Inpainting/Outpainting with SDXL

DanO..

It won't let me vote. I'd like LLM text videos. Which ones are currently the best. I can't get Image Prompt IP Adapter to work. Would be nice to have a look at some problems people might have.

John

Advanced Controlnet, Checkpoint merger, How to transform audio into another person's voice, LLM updates

proteus

do you know of any way to programmatically use the vision model that ChatGPT uses? is there API access somewhere?

Robert Bersch

My current issue is getting xformers to work so I vote for that..lol

robbiek

I think a weekly or bi-weekly 5 minute recap of news related to the three major AI gen fields (image, text, audio) would be great. There's so much happening that it's really difficult to keep up with changes that would actually matter.

Joel Champ

Training for the perfect hands!

Sugar Coat VFX Design

Your Runpod training is very very good. But I am Runpod, Lunix beginner. It will be great if you can share general tips about Runpod, jupyterlab notebook command, etc. For example, I can't see the files inside "checkpoints" folder. I encounter many these small issue when I use Runpod.

eyal giron

Anime/Cartoon Lora Training, as the character Lora Training is based on the site Star By Face that don't include Anime/Cartoon faces

Natan Yakobs

They will add this functionality to API in the future, probably in November

ZenzeosArt

For other LORA training: I would be interesting to learn how to do LORAs like the wowifier or enhance detail ones. That just add detail to any kind of images.

Jovann LEBLANC

I’d like a fine tuning Lora files Tutorial. What parameters I need to tune to obtain lighter weight Lora is an example. Thanks to your previous tutorials, I’ve made stunning Loras for my own needs, but all of them are about 1.7Gb

Tangbohu

I’d like a fine tuning Lora files Tutorial ,finetune SDXL ckpt

Aitrepreneur

same principle as style training, already did a video on on that

Aitrepreneur

the size of a lora depends on network rank, decrease that value to have smaller lora, you can also "resize" a lora inside kohya but of course the final model will be a bit worse

Mark Sutherland

I'd like to see a video on Reactor (face swap for automtic 1111) and one of your one click installs with it

Jason Bognacki

Animatediff with control net and or prompt traveling both int AUTO 1111

cool1

It would be good if you do a video on how to train one of the AI local large language models in oobabooga gui, so that it has more info - including one of the 13B ones - and maybe a 7B one too.

Andro.Meta

ComfyUI becomes much easier once you install the node manager and Stable Swarm. With Stable Swarm, it becomes much more like an automatic1111 interface :)

Martin Thoburn

++ on AnimateDifff with Control net.

James Stewart

It's been 6 months since your vid on using audio in Oobabooga (https://www.youtube.com/watch?v=lSIIXpWpVuk) - if there's been any breakthroughs it'd be great, especially recording a voice that can be used by your role playing character. Personally I use A1111 so would like to see that, but understand for others if a Silly Tavern version is also included.

Jovann LEBLANC

Thank you for your reply :) Does decreasing network rank value mean "worse" model as well ? My goal is to obtain Loras as small as such we can find on Civitai or else, 150~300Mb for an SDXL one.

Aitrepreneur

yes it usually makes the lora a bit worse, although sometimes for some lora the difference is not huge, so you kinda need to try it out yourself and compare them and see if the difference in file size is worth it to you

Martin Thoburn

Auto install Oobabooga install.bat would be nice.

Andro.Meta

I think our Automatic1111 users that are having a hard time with comfyUI would benefit from learning about StableSwarm. It installs comfyui and gives a more automatic1111 ui feel to it. Bonus points to show how to reference comfyui to all the models we have installed over in automatic1111 (it's a fairly easy document to edit in the comfyui folder).

Semon Xue

I was recently exploring the replica.com website, and I found that they provide some very convenient solutions for model training. Is it possible to introduce some practices of lora and model training on replicate.com?

Aitrepreneur

I'm don't exactly know what convenient solution they really offer but seing the pricing, I wouldn't say it's worth it. It's not like the runpod approach is harder, especially since we are using a premade training template, setting up a new training usually takes me less than 5 mins once you have done it once, I call that pretty convenient and cheap. Plus once you know how to do that if one day you decide to buy a powerful machine to do it locally you will know how to do it yourself, so you are also learning some new skills

Semon Xue

I agree with you very much. I learned a lot of sdxl lora training techniques through your videos, especially the use of runpod. My preferred cloud training server is runpod. I look at replicate.com with a new skill mentality.

cool1

Another idea for a video is to show how we can do inpainting with SDXL in the automatic1111 web UI eg. do where do we download the inpainting model for it, do we need controlnet for it and an inpainting model for that (and SDXL one). Also are any inpainting models allowed for use other than research only?

Superluis25

Do you know an IA model like GPT4 x Alpaca or Vicunia which can run on a NVIDIA RTX 3070 with 8GB of VRAM?

cool1

You could check on huggingface.co for the TheBloke's quantized versions of models like the Mistral-7B ones and check the model card. One site says one of one of those - or one of the quantized versions of one is around 4 GB VRAM (it's around 4.3GB file size so it might be). If you have plenty of RAM I think you could also use the ones that use RAM rather than VRAM but I think it will be a _lot_ slower.

Quebber

I'd love to see a guide for some way to do offline text to speech in relation to the WEBUI interface.

Aitrepreneur

What a coincidence, that video is coming tommorow ;)

Oh Gott

I would be very interested in a LORA Training for Products. I just saw the tutorial for a simple beer mug, but I would be interested in building a LORA of photo cameras and equipment to create marketing images of it. THX

Aitrepreneur

The principle is the same, you can train any object with that, including cameras or anything else, it's never gonna be 100% perfect with LORA but you can still make some cool images

Oh Gott

Thanks for your quick feedback. What would you suggest if I want to archive an accurate result? Perhaps increasing the dataset? Is this even possible?

Aitrepreneur

yes the more images of your product you have the better it is, still have some limit, it will never pick up on very small details but with a bit of photoshop and inpainting you can get some pretty cool results

Brad Taylor

many great suggestions already especially the comfyui tutorial, controlnet tutorial (both automatic1111 and comfyui), and animatediff ideas. My addition would be a tutorial on best ways to change a style of an image but without changing composition, etc ie changing a photograph image of a scene to a cartoon of that same exact scene.

pib boris

I m a beginner. I'm a french nurse who started to use IA for qualitative research. La Sorbonne didn't use it for transcribe speach to text and analyse them with IA. I want to learn, i want to make nurds usefull for publique health. It's been 6 mounth i learn a LOT with you and but french ones, on utube are not so good, not one youtube. Sorry for my English.

Peruvian Dota

AnimeDiff/ControlNet to make animations, training of Lora's for objects, graphic organizers or presentation templates (inforgraphics); text-generation finetuning or talking to data offline using vectors/embeddings

Erik Simonetti

A video on training LORA for LLM would be cool.

Nicholas Canada

I want uncensored locally hosted art generator, comparable to DallE3

Eugene Wozniak

Lora training for a specific outfit, for stockings please. I want to make some beautify artl, and want to train different types of women stockings.

Xodroc

I'm interested in training Motion LoRas for use with AnimateDiff

JamesJeffterton

I'd love to see a LORA trained to make unique logos

D W

I am brand new to the Diffusion Party and I Got here thanks to your Video on Youtube. The issue is, because I am clueless, I am having a hard time understanding the difference between Checkpoints and Lora model add-ons like cloths and stuff. I understand that these addons are called "Diffusers"? Anyway the issue is I see some videos saying to put the Lora Checkpoints in a Folder Called Lora and then to everything that is not a checkpoint into the standard models folder, I got very confused and so download everything and then copy everything into both folders just in case. As I said CLUELESS. I cant find any good videos addressing these issues.

Aitrepreneur

It's difficult for beginners, I'm writing a course right now about stable diffusion but it's not gonna come out soon so in the meantime a checkpoint or "model" is THE text to image ai model that is used to generate images, there is a variety of them either officially created by the company stability AI("version 1.5" - "version 2.1" & "SDXL" are the most famous) or trained by the community (you can find them on a site called https://civitai.com/) they all have a different style and feel to them trained on different version of the official release (1.5, 2.1 or sdxl). A LoRA as I explained in my videos, is a small file that basically tells a model what something is supposed to look like, which guides the model into generating images following the training data provided by the LoRA file. You can also find plenty of LoRA files on the civitai website. The checkpoints or "models" are supposed to go into the models\Stable-diffusion folder while the LoRA files are supposed to go into the models\Lora folder. Then you launch the webui and select the checkpoint (top left of the screen or middle of the screen). And for the LoRA file you select them in the LoRA tab located in the middle of the screen (Generation-Textual Inversion-Hypernetworks-Checkpoints-Lora) then use them inside your prompt using the usually provided trigger word.

D W

Perfect expanation!

ragh

Hello AItrepreneur ! Personally, I would love to see a lora training video for poses !

Reinhard Kepplinger

hi, thank you i'd love to see a 360 panorama tutorial upscale and export with multuiple exposures to make a 360 .hdr for lighting in 3d apps

leion800

unstable diffusion, grapefruit if you want anime get stable defusion and the models and boom.

Nipsey Hussle

ai video avatar generated locally on pc. wondershare/heygen

Capt_Bullship

My vote went to Another LORA TRAINING session that would be set in the Warhammer 40k universe and maybe more precisely the whole Adepta Sororitas division.

Kyle Johnson

A tutorial to train and assemble a group of individuals, like friends or the actual TV cast of F.R.I.E.N.D.S , to recreate scenes such as superhero scenarios, or insert them into other TV shows like Stranger Things, or even have them perform as BTS on stage?