AI Updates: ComfyUI Custom Interfaces Studio, OpenAI Sora, Google Gemini

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
hey everyone welcome to another video what if I were to give you this particular FR and you get this output out is this good enough in 2024 or maybe you get this one or maybe this one or maybe you're going to go out and grab your awesome workflow with 200 plus NES and then you come up with a full key high resolution I need detail image and what if I tell you that this bu key super highly detailed image is not good enough this is open AI Sora a text to video model and look at the detail look at the detail the skin the reflection this is absolutely amazing Sora is an AI model and it's advertised as creat realistic imaginative scenes and goes on it's even text video model it can generate video up to 1 minute long Main maintaining the quality and the consistency and you can see this was the video I showed you earlier we have the front there front goes into the AI outcomes a one minute video as another example this one is about mammoth in snowy terrain one more example this one is about a movie trailer so you understand what is a movie trailer and if you were to look at this you would think that this is in TR shot a couple more example this one is papercraft underwater real life fish and sea creation can also do portrait mode but if we look carefully if you know what we are looking for you can see that some of the frames they are not realistic you look at the white ship you can see there but if you were to see these videos on Instagram Twitter if you do not know what you're looking for you totally think that it's real it's made by people and it will only get better if we take Chang it will only get better and better and this one is absolutely beautiful if I were to see this on Twitter I would think that this is from a flagship game release in 2024 this is really difficult to generate from an AI we have the camera following the SUV now if you look at the website talks about the safety and it talks a little bit about how DET text classifier rejects input if we have these type of content in there also goes into the research paper telling you a little bit how it works so if you're interested I will have a link in the description where you can read more on this however one thing that open AI does not do is open source the model they always go about security and trying to control how people are using the model let's take a look Del chbt they are awesome tools but we don't have access to the model S du will however is releasing G n i 1.5 and they call it 1.5 Pro where developers will get access to the model in production so they will be able to use it in production and this one can take up to 1 million tokens I'm not sure in they are exaggerating the numbers here but we've seen 128,000 tokens 200,000 tokens but now 1 million tokens and from a large language model this is really good if it can deliver and this one was Gemini 1.5 next let's come back to config Y and a couple of comfyi skin or launchers this one is comfyi launcher it's basically running comyy workl without any setup it's advertised as without zero setup if you scroll down we can see a couple of features basically automatically installing custom nodes the we have workflow exports can open multiple config y we flows and so on there a video demonstrating how box it's based on comy y manager if if you scroll down you will see that it gives credit to comy Y manager to install those missing custom notes pretty similar you can see a comfy y launcher pretty much like comy y but multiple work Flows at the same time there's instruction on how to use Docker or manual installation and the inside is quite simple just get clone and started and here's the credit to confy UI manager so this one was confy UI launcher now next we have confy studio and the name is a little bit confusing because it says confy y Studio but if you check the description it's basically just an improvement on how to manage models so it can categorize the different models based on tags we have the workflows and these are the models and then we can use that workflow in our coffee UI and you can see there's a Luder model manager thumnail shielding all there naughty models automatic model labeling it can also match worklow with model or model with workflow and then support Chinese as well as English there's instruction on how to use this studio super easy keep in mind that it works with couple of supported loers is the list and here's another one con it's conflu space I always confuse the name with confy space but this one is conflu space and it's probably the most visually appealing skin or conf it's again an open source AI front n or confy and it gives a better interactive experience than the default comp one take a look at this screen this is basically what it does we have the workflow the we can explore the different images that we' generated and here's an example of how it looks so you can see we have a simple case ample here and the buttons the drop down is better in terms of The Styling is the positive up load control net model we have the input image even the empty latent image is visually better than the default confy one at the top we can see that we have multiple work flows arranged in different tabs so this was conflu space I gave it a try it work with simple workflows but the moment I had complex workflows with recent ndes it crash this is still in work in progress if you'd like to try it it's a valuable give it a try and maybe report any bugs that you find now next a couple of custom news by the community and this one is comp V UI portra Master it's main Miss is to generate human portrait using these sliders and basically what it's doing it's whenever you add or increase a particular slider it will add that slider to your prompt so it will add dry face 0.5 to your promp as a postive and there are a couple of example you can see it's pretty straightforward to use so instead of you typing small eyes you can just select it from the slider there and here's another one by zoo zoo zoo and again he's going to call it as The Unofficial implementation basically is just a wrapper and this one is you the world the efficient Sam and it's a masking technique that can detect different subjects and you can see here in the example and basically it's going to create a MK for each individual subject there we go so we have a couple of percentage and this goes into the mask here I'm scrolling down it's called as the and mass separation feature capable of individually outputting select Ed mask and there's an example here a little bit down so in this case we just have a general mask identifying the subject here and here's a good example of what this can do so if I zoom in here you can see that it's looking for hair and jacket as two things that we want to select or mask and then our default one so if you have the comy UI impact back basically pretty much extend all now when it comes to masking you will get the B box detector which can detect different subject and in this case you can see on the right side it's detecting the face as well as the hair and then jacket as the whole body here but if we go and check the efficiency sound the Y version you can see that it's detecting very closely the jacket and just the hair not the face so basically it's a a better way of masking we'll have to experiment to see how well it does next we have something related to 3D and this one it's called The Comfy UI 3D pack it's an extensive Noe to working with 3D inputs measures and UV texture and there's a good example down below keep in mind that it is a work in progress let's check the video and we can see here that it's a confy wild world flow you just have the noes that is providing and the output is a 3D object that can be move around so I'm guessing this can be exported to maybe Unity or G game engine can be use there or maybe even blender and here we can see that it can save and load 3D file as obj and blender can open a BJ all right so if you stayed until the very end of this video thank you you're awesome I will leave you with a bonus and I think this one is the most underrated comyy custom nude that we have so far if you are just starting out with comfyi I think this should be your number two custom new that you installed number one of course being the cpy manager and this one is confy assistant is by open art the place where we upload workflows it job is very simple it's going to give you the search inside of comyy where you can look for work flows and surprisingly it only has or stops to install it it's very easy just search for comy UI assistant and install it from the comy UI manager let me give you an example of how to use it so here I am with the N install I have a default workflow and at the bottom you will see that I have a small magnifying glass click on it it says search for workflows and type in pixel let's look for a couple of pixel art workflow and it gives a popup right inside of confy UI where I can choose the workflow you will see on the right side I have use workflow again click on it saying loading and it loads the work through right into config y I don't have to go download the worklow and load the workflow inside of conf one I have all the notes everything is there now if you want to use it to study and then build your own workflow on top of that great you have the workflow you can do whatever you want with it or if you don't like it you can search for a new one in this case let's try Laura let's look for a worklow that talks about Laura basic one and in this one we have a n that's great at bonus telling us how to use Laura now let's look for something a little bit more complex I'm going to choose this simple is workflow again click on use workflow and here you can see that it's telling me that I'm missing custom node click on close click on manager just behind me and then click on install missing custom node then install all the missing custom node and simply restart the application and you will get your workflow back it's very simple it's down there it does not take a lot of system requirements basically just in at this point where you can search for works right inside of confy all right guys so thank you for watching until the very end I would like to say a big thank you to all of you supporting the channel in any way that you can whether you're liking the video commenting out supporting on patreon or sharing the the video I really appreciate all the support thank you thank you to all have a nice day I will see you in the [Music] next
Info
Channel: Code Crafters Corner
Views: 397
Rating: undefined out of 5
Keywords:
Id: njuG70D4MzE
Channel Id: undefined
Length: 12min 50sec (770 seconds)
Published: Wed Feb 28 2024
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.