This AI Deepfake Went Viral - This is How to Make it

[@Bsprouts]

once again, another banger video! Appreciate you taking the time to put these together!

[@digital_magic]

Hey there and thanks for your nice comment

[@MrNorBro]

Nice stuff!!! :D Just what i was looking for! Keep going! Cool Chanel! ;)

[@digital_magic]

Thanks! 😀

[@BK8GamePhSlotTrick-kq9lf]

what is the system requirement of the diffusion Deform? I want to install its pleas answer sir

[@aminshallwani9369]

Hello I have seen you other video also but when hit Generate got this error Error: 'NoneType' object is not iterable. Please, check your schedules/ init values.

[@digital_magic]

I've had many comments like this and most of the times it help if you disable the init image and delete the path in there like I showed in the video. for some reason stable diffusion always creates a path, and this is something I couldn't change in the base deform settings file. please let me know if this helped and I wish you a very nice weekend

[@aminshallwani9369]

@@digital_magic Thanks will check this

[@unreal_unit]

Another great tutorial, thanks 🙏

[@digital_magic]

Hey Andrey, thanks for your nice comment :-) But the only reason this tutorial became like it is, is all because of you. I really appreciate your immense help and your great knowledge. wish you a very nice day

[@AIWithCarl]

What a great tutorial, thanks a ton my dude! Subscribed.

[@digital_magic]

Thanx and glad you liked it 🙂

[@rochgtv]

Do you know where I can get the prompts? They recommended that I take them from Civitai for hybrid videos but I don't know where they are to copy and paste a model that I liked. Or do you know of any page where I can get the prompts I need for deforum?

[@digital_magic]

if you use loras then the prompting is easy. If you dont use loras then try a lot and keep changing or removing stuff from the final prompt

[@GameBoy-mj6fg]

Is using temporal net better than using open pose?

[@digital_magic]

you can use both, but temporalnet gives consistency :-)

[@lxic-bz8hf]

great work, how long it took to render ?

[@digital_magic]

Thanks I'm glad you like the video, my video took about 6 hours to render which is massive. I am working with a 8 GB vram card

[@yuta6273]

Where is the Deliberate model? The link doesn't work anymore. It's gone?

[@digital_magic]

they have an update i guess, just type deliberate in civit ai or try a different more updated model :-)

[@Onur.Koeroglu]

Yea man... Cool.. This is what I needed.. 😎✌🏻 I like that you also describe from where you had the Infos, and from where I could take / buy more Infos.. But first I have to make the Tut "Unlock the Ai... " 💪🏻🫡

[@digital_magic]

Hey there and thanks for your nice comment, I am glad you liked the video. And for sure I need to tell the people where I got my infos from because this is the only way I get better and we can all grow in the community. good luck with all the tutorials and I hope to stay in touch Hey there and thanks for your nice comment, I am glad you liked the video. And for sure I need to tell the people where I got my infos from because this is the only way I get better and we can all grow in the community. good luck with all the tutorials and I hope to stay in touch

[@twentysixSOLO]

Super great beautiful tutorial

[@digital_magic]

thank you 🙂

[@DustyDiffusion]

Hey, great tutorial again! Thanks. I just wanna ask what exactly meaning of the numbers you put before the prompts. Are they seconds or starting point of frames?

[@digital_magic]

Hey there and thanks for your nice comment I am glad you liked the video. yes the numbers I put before the prompts are the seconds of starting points of the frames where the new character starts. I hope this helps

[@timkpk]

Would you tell me why I cannot found controlnet control type tile? I just found tile/blur instead, is it the same?

[@digital_magic]

no, you have to download it from the huggingface site, or/and reinstal the controlnet extension

[@igotthegoodness]

can this be done with out SD locally installed?

[@digital_magic]

yeah you could use something like rundiffusion

[@muggle_world]

thanks for this tutorial man...i will try this in my next video..changing weights of tile and temporalnet really do wonders i created goku with weight shifting and 3 cn..it was good

[@digital_magic]

Hey man thanks for your comment I'm glad you liked it. you guys any links to your video? would love to see it. hope to stay in touch

[@muggle_world]

@@digital_magic its on My channel last 3-4 videos made with CN weight shifting

[@marilynlucas5128]

You’re doing great work.

[@digital_magic]

Hey there and thanks for your nice comment, I am glad you liked the video.

[@AIExplor.e]

When I click on generate and im looking at the folder output to see the results, it's completely something else. First image its my image after, it shows me completely something else. I do exactly as described in your video

[@digital_magic]

have you tried disabling all big extensions and re-install controlnet and deforum?

[@digital_magic]

In my unlocked the AI video potential in stable diffusion video I explained all about this, that you have to disable big extensions cuz otherwise it's conflicts with controlnet and deforum

[@digitalflick]

guys dont spend your money on settings, there is no catch all solution. every video is different. spend a week experimenting.

[@digital_magic]

Hey there and thanks for your comment, I think you are absolutely right with that, that every video needs a different solution and it cost a lot of experimenting to figure out what works best. I think that we all can learn from each other. I am just very grateful to Andrey from unreal unit and this is how I wanted to share my appreciation to him. what I said in my video he shares his knowledge with the community and if there are people that can afford it to support him then I think that would be more than fair. I wish you a very nice day

[@erdbeerbus]

great, again ... thumbs up!!! but when I set the denoising strengh to 1 it is going to be crazy results, ot the img2img image ... grz!

[@digital_magic]

Have you set the img2img noise multiplier to 0.05

[@erdbeerbus]

good advice ... will try it //write the answer here asap

[@blueyc4rter]

Do you need to keep the seed the same in img2img and deforum? How else do you ensure the deforum output follows the img2img, or is it just about finding a good prompt?

[@digital_magic]

In deforum, You always set the fixed seed to -1 you can't ensure that the output you have created in image to image will be the same in deforum. It is exactly what you said in the comment it is all about finding a good prompt and then in deforum I always give it four or five and sometimes even 10 tries till I get the desired result when clicking on the generate button. hope this helps and I wish you a very nice day

[@mvsell9037]

your videos sends me to meditative state. feels like im meditating

[@digital_magic]

Hahahahahha, I hope this is a positive thing for you? I don't hope you find my videos boring? I tried to make the beginnings of the video as entertaining as possible to motivate the people to watch the video. then there's always this Long Bit, where I have to explain everything and which can be boring and meditative as you said 🙂

[@SimonWackerle]

This is for sure a compliment! I feel the same...very relaxing 😊

[@digital_magic]

thanx bro @@SimonWackerle

[@yvann_ba]

Thanks a lots for your videos

[@digital_magic]

you are welcome 🙂

[@rightside8937]

Thanks a lot...but I can't see the stable diffusion checkpoint menu to choose my model... :(

[@digital_magic]

have you tried to update stable diffusion?

[@rightside8937]

Thanks! Yes now it works, I have the model option with 1.5 pruned ..I don't have noise multiplier but I will reinstall it @@digital_magic

[@Damian_57]

deforum，Can you control the size of the redraw?Thank you, sir.

[@digital_magic]

sorry i dont know what you mean?

[@ai_and_gaming]

This is pretty cool, never seen before

[@digital_magic]

Hey there and thanks for your nice comment, I am glad you liked the video.

[@Editboyka]

Can you send me the link for website please

[@digital_magic]

which website?

[@alishkaBey]

great tutorial!

[@digital_magic]

Hey there and thanks for your nice comment, I am glad you liked the video.

[@neocaron87]

It looks awesome, I've been experimenting with my 4090 with Sdxl and the latest controlnet using diffusion temporal model combine with control loras which runs very fast and are not memory hungry (you should try!) and the consistency is through the roof!

[@Onur.Koeroglu]

Hmm.. Do you have me a Suggestion or link for that? 🙃 I'm always interested in consistency.. 😁✌🏻

[@muggle_world]

@@Onur.Koeroglu more weight you give to tile more consistency you will get and try to play with cfg settings too

[@digital_magic]

thanks for your nice comment, this sounds awesome what you are saying about a memory which is not hungry :-) do you have any tutorial links or any articles which I can read so I can learn more about this? I would really appreciate it. wish you a very nice day

[@neocaron87]

@digital_magic I'll share the links regarding controlling LoRa in my next comment. If you don't see them, it's possible that KZbin flagged my comment as spam because of the link. To provide a quick explanation, the creator behind the Temporelnet model has released a new model for SD XL. As I mentioned earlier, I typically avoid using Deforum with it. Instead, I combine batch images with the temporal model and utilize two controllora parameters (in controlnet): depth and canny. There are both 256MB and 128MB versions available, and the latter might work better on your setup. Alternatively, you can use the low VRAM option to address any resource constraints. The key step here is to convert your video to either 4 or 8 frames per second before initiating the batch rendering transformation. Once that's done, you can employ Topaz Video and the Apollo interpolation feature, which creates additional frames based on your existing ones stabilizing and blending the frame in a nice looking way. Surprisingly, it performs exceptionally well. A recent discovery of mine is that you can simultaneously deflicker the video using the enhancement option in Topaz and playing around with the setting (my recommandation is all the slider at 100% except of noise reduction at 0 and deblur at 0 too). Once I'm entirely satisfied with the results, I plan to create a video detailing my workflow. Hopefully, it will prove helpful, even if you don't have 24GB of VRAM! 😄

[@digital_magic]

@@neocaron87 Hey there and thanks for your amazing comment. you are right I didn't got any message with any links, so KZbin probably blocked it. thanks for your amazing information it sounds very interesting and I'll be looking forward to seeing your tutorial. maybe we could get in contact through Twitter Instagram or tiktok, if you have any of those accounts? my addresses are: Instagram: digital_magic_1 Twitter: @Digital_Magic_ Tik tok: digital_magic_fan Looking forward to chat with you on one of these platforms as you seem to know a lot about stable diffusion and will be great to share our knowledge. wish you a very nice weekend

[@AIEnthusiasm]

thanks for this video

[@digital_magic]

Most welcome

[@Mundagreen]

Thanks for your vidio, from indonesia

[@digital_magic]

glad you liked it 🙂

[@roofcleaningsolutionsorlan6595]

hi thank you for great tutorial...i did everything that was mentioned in video but at end i got this error Error: 'NoneType' object is not iterable. Please, check your schedules/ init values. ...can you please help solve it

[@Onur.Koeroglu]

Hey I had this dumb failure yesterday too.. I tried all.. Reinstalled ControlNet, Deforum and in than Automatic 1111 with no success.. The only thing that helped me was to reinstall git bash and python... 😑😐 When you find another way, ao please let it me know..

[@digital_magic]

I've had many comments like this and most of the times it help if you disable the init image and delete the path in there like I showed in the video. for some reason stable diffusion always creates a path, and this is something I couldn't change in the base deform settings file. please let me know if this helped

[@digital_magic]

@@Onur.Koeroglu Thanks for your reply, I haven't heard of this before. what I mostly do if I got problems I go to the Discord group from deforum

[@donttreadonme154]

Where the Andrew Tate version

[@digital_magic]

Hey there I'm not sure what you mean?

[@donttreadonme154]

@@digital_magic I would have Andrew Tate dancing on a Bugatti before the scene transitions into prison cell while he smokes a cigar, with maybe Greta delivering a box of Pizza in the background

[@digital_magic]

@@donttreadonme154 That sounds like a very funny idea for a video. I'm not sure if I'll be able to create something like that. but thanks for your great input. I wish you a nice day