yes I watched 20min, if you don't know where you came from its hard to tell where to go moving forward.. your research and recap was amazing !!! thank you so much..
Thnx, i just started, downloaded bunch of models and tryed them out, then my hard drive space vanished and i wanted to understand what is what. So i stumbled on this video and it more cleareer to me and now i will ignore SD 1-2 and find few models from SD 1.5 XL to play with. and save Disk space =) Thnx again.
thank you this was great! I really needed summary of the history. made me feel better I'm still stuck with mostly sd1.5 and just now a bit xl. not missing out on the others it seems
I went from 1.5 SD to PNY. Skipped Standard XL. PNY's just better despite my initial reluctance, More accurate prompting, better hands and anatomy, while remaining somewhat compatible with existing SDXL loras. Have not returned to 1.5 for months because of it.
Man thx for this it thought me alot I didn't know stabilityai ran out of money 😢 I hope they can get back on there feet maybe they can make a new model that we have to pay for like midjourney
Looks like lately they've made some deals to host Stable Diffusion 3 Large on Amazon Bedrock (the open model we can use is SD3 Medium), so yeah it looks like they are doing a kind of premium paid model tier.
Interesting to learn some of the history as I only got into SD earlier this year. SD3 seems a dead end to me, it'll be Flux going forward until the next big thing.
Yeah for the most part Flux is very similar to Stable Diffusion to the point where it's supported in SD front ends like ComfyUI and Forge. One thing I didn't mention is the text encoder, which allows Flux and SD3 to generate accurate text in a way that the architecture just doesn't support in SD1.5 or SDXL, so that's one major difference.
nice video. I was trying to make the point on stable diffusion, and its well done. During your video, you show a graphic with different version. Do you have put it or take it, somewhere ?
Glad you like it. I made the timeline graphics myself. I'll try to link them and hope youtube doesn't delete my comment. files.catbox.moe/n8eo85.png files.catbox.moe/fzc3q2.png files.catbox.moe/6yqnpp.png files.catbox.moe/z7vzl2.png
@@ChameleonAI thank you. And dont worry, as teacher, i know what its look like to done something nobody care. But in fact, someway, i suppose its matter ;)
You'll want to look into training a LoRA. For best compatibility with today's most popular models, yeah I recommend training that LoRA on XL 1.0. If you want to be very cutting edge you might look into Flux.1 which is outside the scope of this video but the LoRA training process is similar from what I've heard. (I'm not a LoRA training expert so I can only give you general advice)
yes I watched 20min, if you don't know where you came from its hard to tell where to go moving forward.. your research and recap was amazing !!! thank you so much..
Glad to see Martin Sheen took time out of his busy schedule to narrate this.
Really enjoyed this video! You did an awesome job with the research, and your voice is super nice. Keep up the great work!
What a great summary and research, well done
Thank You! this completed my knowledge of SD.
You're welcome and thanks for this great overview!
Thnx, i just started, downloaded bunch of models and tryed them out, then my hard drive space vanished and i wanted to understand what is what. So i stumbled on this video and it more cleareer to me and now i will ignore SD 1-2 and find few models from SD 1.5 XL to play with. and save Disk space =) Thnx again.
Great video, I still finetubing my own sd 1.5 models and loras… ty
A short trip down memory lane for me! If you know where that first asuka picture came from, you know how everything started!
It's a shame the original voldy guide doesn't exist anymore. The Asuka picture was a big deal. If you know, you know!
thank you this was great!
I really needed summary of the history. made me feel better I'm still stuck with mostly sd1.5 and just now a bit xl. not missing out on the others it seems
I went from 1.5 SD to PNY. Skipped Standard XL. PNY's just better despite my initial reluctance, More accurate prompting, better hands and anatomy, while remaining somewhat compatible with existing SDXL loras. Have not returned to 1.5 for months because of it.
Thank you, fascinating
thanks . great video. lot of information
Man thx for this it thought me alot I didn't know stabilityai ran out of money 😢 I hope they can get back on there feet maybe they can make a new model that we have to pay for like midjourney
Looks like lately they've made some deals to host Stable Diffusion 3 Large on Amazon Bedrock (the open model we can use is SD3 Medium), so yeah it looks like they are doing a kind of premium paid model tier.
thanks grandpa 😍😍🔥🔥🔥🔥
Interesting to learn some of the history as I only got into SD earlier this year. SD3 seems a dead end to me, it'll be Flux going forward until the next big thing.
Yep I agree. Since publishing this video, Flux has already become more widely adopted than SD3.
the flux architecture may be different, but it generates the same thing as the previous models
Yeah for the most part Flux is very similar to Stable Diffusion to the point where it's supported in SD front ends like ComfyUI and Forge. One thing I didn't mention is the text encoder, which allows Flux and SD3 to generate accurate text in a way that the architecture just doesn't support in SD1.5 or SDXL, so that's one major difference.
@@ChameleonAI I think if add 50 lore to a pony and the ability to understand descriptive or poetic language will be like flux
nice video. I was trying to make the point on stable diffusion, and its well done. During your video, you show a graphic with different version. Do you have put it or take it, somewhere ?
Glad you like it. I made the timeline graphics myself. I'll try to link them and hope youtube doesn't delete my comment.
files.catbox.moe/n8eo85.png
files.catbox.moe/fzc3q2.png
files.catbox.moe/6yqnpp.png
files.catbox.moe/z7vzl2.png
@@ChameleonAI thank you. And dont worry, as teacher, i know what its look like to done something nobody care. But in fact, someway, i suppose its matter ;)
i always said, advancement in technology is either because people are horny or need money. SD timeline is the pinnacle of that saying
comment for the algorithm gods
Are fooocus and stbale diffusion use the same nodels?
Yes, Fooocus is based on Stable Diffusion XL
Yes, Fooocus, A1111, Comfyui are frontends and Stable diffusion is an architecture, and juggernaut, realvis, epicrealism are models
that was interesting until the end! would you use xl if i want to finetune my own face ? im new and still figuring it out
You'll want to look into training a LoRA. For best compatibility with today's most popular models, yeah I recommend training that LoRA on XL 1.0. If you want to be very cutting edge you might look into Flux.1 which is outside the scope of this video but the LoRA training process is similar from what I've heard. (I'm not a LoRA training expert so I can only give you general advice)
Are these comments AI generated? Am I AI generated?
Am I stable? Am I diffused?
I just don't know...
Maybe we're all just hallucinations of an AI.
And this Is narrated by using a text to speech software...
It's speech to speech. It's me talking into the microphone, I just transform it to a different voice.