unfortunately the lighting of the character is still inconsistent with the background lighting but it's really exciting to see how the field progresses in image composition. As much as i like comfyui or Forge, the gui should really move towards something more akin to image composition and less cable noodle management or text box madness.
Thanks for the great video, Olivio. That use of coloured box to give more ambient colour takes me back to your Affinity Photo tutorial days. This is not a paid comment.
The 3D model conversion IS pretty cool. However I refuse to pay for these kinds of services that just piggyback off of publicly available tech (Flux in this case) and then keep 'their' work proprietary.
Hi Olivio! I'm not sure if you've discussed this already, but I was wondering if you had a video comparison explaining the pros and cons of going through virtual machines like Runcomfy, vs locally running Stable Diffusion. I prefer to host it locally, but my computer is far slower than what they have available on Runcomfy, especially for video generation. I'm trying to decide on whether to invest in a new beefy PC with the upcoming GPUs with AI integration, or just rent via Runcomfy. I'm sure there are others trying to go through these options as well, so I figured a video from you covering it could be helpful! 😊 Here's my thoughts so far while trying to decide: Locally: Pros - Offline - More privacy - Full control of files, but also can be frustrating and time consuming to setup and manage - No need to upload assets to a server before using them - Can control it from my laptop if I really wanted to, so that make's Runcomfy's flexibility obsolete. - Lots of other benefits aside from image and video gen with the new GPU, and I game a lot Cons - Limited in processing power compared to the many machine tier options in Runcomfy - May be unable to do as much while processing - Huge upfront cost, which will potentially devalue much quicker than usual, due to rapid AI advancement, so higher risk Runcomfy: Pros - Zero frustration or wasted time with setup or management, plus they have support if you really need it - Can switch between machines easily, and the beefier machines will pump out videos in a fraction of a time, saving many hours - Easy to collab if you wanted to - Very small upfront cost compared to a new rig - Can work/do whatever else on your pc while generating on Runcomfy without losing any processing power. Cons - Cost over time, if you're needing the nicer rigs for hours per day, that could be up to 3k a year, so you might as well buy a new one or parts. - Have to upload and download assets plus manage their files on their cloud. It's probably best to just get the new parts, and use Runcomfy sparingly when you really need it, or use it until you can afford the new rig.
As always, great video… but… where are your glasses ? I am also wearing glasses. What is your quick advice for training models (do i remove my glasses and use a prompt to add them later ?).Thank you
i took my glasses off for the training photos as you can see. you can also make photos with glasses, but maybe make sure you have not much reflections in them. also, unlike me, try to wear different clothing in the photos. because my lora is pretty much trained on a flower hawaii shirt
Bro this is perfect for making consistent scene for AI video, using our own visuals of us or some other real world character types. Dope! Does Krea charge to use this new Lora real time training thingy? Like can we test it for free
This is NOT a paid ad 😎👍
unfortunately the lighting of the character is still inconsistent with the background lighting but it's really exciting to see how the field progresses in image composition. As much as i like comfyui or Forge, the gui should really move towards something more akin to image composition and less cable noodle management or text box madness.
they should pay you
Thanks for the great video, Olivio. That use of coloured box to give more ambient colour takes me back to your Affinity Photo tutorial days.
This is not a paid comment.
Thanks for sharing this! Now all we need is a comfy workflow similar to this with realtime flux. 🙌
I will have a look at that. It's high time I do comfyy videos again
@OlivioSarikas 💯
The 3D model conversion IS pretty cool. However I refuse to pay for these kinds of services that just piggyback off of publicly available tech (Flux in this case) and then keep 'their' work proprietary.
Thanks, saves me watching this video knowing its a paid thing.
Thank you for your video. Great.
Hi Olivio!
I'm not sure if you've discussed this already, but I was wondering if you had a video comparison explaining the pros and cons of going through virtual machines like Runcomfy, vs locally running Stable Diffusion.
I prefer to host it locally, but my computer is far slower than what they have available on Runcomfy, especially for video generation. I'm trying to decide on whether to invest in a new beefy PC with the upcoming GPUs with AI integration, or just rent via Runcomfy. I'm sure there are others trying to go through these options as well, so I figured a video from you covering it could be helpful! 😊
Here's my thoughts so far while trying to decide:
Locally:
Pros
- Offline
- More privacy
- Full control of files, but also can be frustrating and time consuming to setup and manage
- No need to upload assets to a server before using them
- Can control it from my laptop if I really wanted to, so that make's Runcomfy's flexibility obsolete.
- Lots of other benefits aside from image and video gen with the new GPU, and I game a lot
Cons
- Limited in processing power compared to the many machine tier options in Runcomfy
- May be unable to do as much while processing
- Huge upfront cost, which will potentially devalue much quicker than usual, due to rapid AI advancement, so higher risk
Runcomfy:
Pros
- Zero frustration or wasted time with setup or management, plus they have support if you really need it
- Can switch between machines easily, and the beefier machines will pump out videos in a fraction of a time, saving many hours
- Easy to collab if you wanted to
- Very small upfront cost compared to a new rig
- Can work/do whatever else on your pc while generating on Runcomfy without losing any processing power.
Cons
- Cost over time, if you're needing the nicer rigs for hours per day, that could be up to 3k a year, so you might as well buy a new one or parts.
- Have to upload and download assets plus manage their files on their cloud.
It's probably best to just get the new parts, and use Runcomfy sparingly when you really need it, or use it until you can afford the new rig.
i want to bring that out real soon
fantastic!! this is exactly what i was looking for.
Its great, right? Amazing how far Ai has come
Do you have any guarantee they won't use your images or loras made of yourself to train their future models?
Why would they do that and why would I have a problem with that?
@@OlivioSarikas replicate you and make new videos from you on a different channel. And make profit with voice clone... And many different ideas..
I love your videos. Thank you so much for making them.
Thank you, that means a lot
so good
As always, great video… but… where are your glasses ? I am also wearing glasses. What is your quick advice for training models (do i remove my glasses and use a prompt to add them later ?).Thank you
i took my glasses off for the training photos as you can see. you can also make photos with glasses, but maybe make sure you have not much reflections in them. also, unlike me, try to wear different clothing in the photos. because my lora is pretty much trained on a flower hawaii shirt
Bro this is perfect for making consistent scene for AI video, using our own visuals of us or some other real world character types. Dope!
Does Krea charge to use this new Lora real time training thingy? Like can we test it for free
the lora training costs a little bit, but it is included in the subscription by using credits
So awesome!
yes, i love it too
What can be made as a gift for Valentine's Day? Urgent help
roses and chocolate always seem to work nicely ;)
Awesome beards all round!!
Thank you. I wish the ladies would like long beards more so I could grow a epic man mane 🙂
Cool vid, but prefer to use local models only
Yew, I would like that too. I have to check if there is anything like that local
Generative AI is getting out of hand! THANK YOU OLIVIO🎖
Thank you, you are welcome
If they had a clothes try-on feature, I would subscribe.
Luckily that is something that can be done quite well locally already.
That's a good idea. I will suggest it to them
cool but the lighting on the character doesnt match the lighting of the scene
you can certainly spend more time to improve that. this is just a quick demo
website? no open source alternative? PASS!
Indeed, this YT channel is always promoting paid stuff lately.
It doesn’t work with long hair and wardrobe, hence there’s no consistency. Another bummer from Krea and the AI industry in general.