The ones you think don't look like you, actually do look like you

Matt judges the similarity to his own face much stricter than other people's faces. All of the faces of Matt are more than good enough for the desired thumbnails. The workflow is so much faster.

Dude, you are so inspiring!!! I truly love your videos, your newsletters, when I get a break from work I watch all your cool stuff. It's the best time to be alive for people like us loving AI and you not only prove that but also add so much value to this subject. I really really appreciate you and your work you do for us. Just wanted to say that.

The face replacements actually look great! You should definitely use this method, to save you time!

✨PROTIP✨ If you include the parameter --iw 2.0 at the end of your MJ prompt when including your own picture as an image reference (aka image-prompt). It can help make the resulting images resemble you twice as much as it possibly can. By default this weight number is at 1.0.... so increasing it to the max of 2.0 does help.

(0:14) That second image of you backpacking through the Kingdom of Arnor is SPOT ON!

If you apply multiple times the INSwapper it actually looks better. I've been doing it and it helps me get closer to my subject. I usually apply 3 times.

THIS IS HOW.A TUTORIAL IS DONE!!! THANK YOU!! You walk us through it, you don't cut to other crap!! Well done. You're a very good instructor. You try to cover as much as possible and you try your best to not miss any details.

I've been using this one for months. Dragging your swapped image into Photoshop and using the liquify filter allows you to perfect the head and face shape.

Oh man! I thought you were the one who told me about this a couple months ago. I actually used it to swap out faces with a bunch of collaborators when we did a writing relay on KZfaq, and we put this on our thumbnails as a marvel superhero team. Thanks so much for sharing this with those who don't know about it!

Our big brains sure like to change things up! Great to hear about resetting and returning to your roots. Love the personal touches! 🙏

Congratulations on 10k in Discord buddy! 😎👊 That's a great place to be growing. Day 1s are your A-1s!

Wow, thank you for this tutorial. I had fun with it all afternoon with excellent results. Your channel is fantastic.

This should help with cconsistent character images for graphic novels!

I’ve been using this for months. It’s great! Now that you’ve shared it, I’m nervous the service will slow down as you have a large audience. 😂

🤯🔥 I was waiting for this!! Thank you

great video, I have been using Insight for about 2 months. It's been a game changer. It's really cool to watch old movies, and grab a face of someone obscure, and swap their face, and then make a lora or embedding of them. It used to be 50 ifree mages a day, no I think it's 21 or so. Still not bad. also it doesn't matter of they have glasses. I did my Mom, the only good image of her when she was young was a portrait with cat glass from the early 60's ,works like a charm.

Matt, I really appreciate your work and videos! I suggested your channel to teachers willing to know how to properly use AI in education. Best wishes from Romania! Adrian

You can also generate an image with multiple people, though 2 or 3 works well, not huge groups. Then you can swap in several saved ids. So for example, I could generate me and 2 friends in front of the Eiffel Tower, then swap in our saved ids putting them in order from left to right, separated by commas. Thanks for another useful video insight Matt.

How hilarious, I’m literally trying to do this as your video comes out

So fun! Love the pics it created. Going to try it out for some thumbs on one of my channels. Thanks for showing us how to use it! :)

Looking forward to more videos like this Matt! And hahaha agreeing with everyone else that you know your face too well to notice but the Inswap images really actually do look like you.

"not much changed about it" 🤣🤣 I mean you are hansome but not as chiseled as the first image lol

Matt I love your videos. You are the best AI youtube channel by far.

Doing a /blend of the Mid Journey image and your source image before using inswapper could improve the results.

The speed and accuracy of face swapping with mid-journey has definitely improved. Matt's meticulousness in judging the similarity to his own face showcases his dedication to perfecting the process.

Been using this for a little while. Best one for this option

Have I missed your process doing your thumbnails??? That would be cool to watch for sure. Love from Brazil and tks for all the great you’re putting out there 🙌

This is how I’m creating all my AI videos for a few months now :) I love it!

You are so awesome!!! Thank you for all you do and sharing so much!!! I learn a ton from you and share your vids like crazy!!!

Oh love all ur updates ❤always look forward to ur videos

Yes the Insight Face Swap works well for photos, it doesn't work for drawings yet though. ALSO it has trouble with extreme mouth positions. Generally it works very well.

thank you for this update Matt! 🙏

Wow... this works and it easy to do. Thank you Matt for sharing and being an awesome human being

Great videos Matt! Thank you!

Awesome I look forward to watching the new content!

I love the idea of this. It needs a lot of work, but they'll get there eventually. It's nice to know it's being developed.

Fire Man! Love The Content! 🔥🔥🔥

Im super proud of you my man! don't let this channel overwhelm you, take things one at a time and don't skimp on family time. We'll be here, waiting, so take your time buddy!

These are the best videos!! Can you make one about your video editing workflow?

Haa LUVED seeing our AI RingMaster as a Ken bro!!

Thanks for the tutorials! Super helpful😊

THANK YOU MY GOD BEEN LOOKING FOR THIS KINDA OF THING FOR SO LING

Great video, thank you Matt👍

You can also inswap an image multiple times. So if you inswap the face that has already been swapped you will get closer

Thank you Matt for bringing the good stuff! 😎

Interesting stuff. Not bad. Going to experiment thanks to your review. Thanks man.

Love this video plz continue to make more like these

Thank you buddy, much love ❤

It's a great tool, a little work on getting the eye colour (they always look dark) right will go along way. I use photoshop, create 2 layers (make image same size), original as bottom layer and with a little work you can use the eraser tool to get some colour in the eyes. I do subscribe to it and worth every penny, has alot of uses.

Appreciate you!🙏🏾

You can use the prompt --iw 2 to indicate the weighting of the source image (2 being max)

oh wow I actually knew about this several weeks before Matt!! lol. In all seriousness it does work pretty well unless you have a really stylized image.

Nice tutorial, thank you!

its like everyday a new expectation. i love this knowledge

Matt, you need to select faces that are closer to the shape of your face. Try using a painterly style, a pencil sketch, or an illustration.

Liked this one definitely Matt🤘

Looks really interesting.. I might have to steal some hours and explore the animation option they mention. 👍

I've been using it for a few months, it works well.

very infromative, thanks for sharing

imagine this tech 6 months from now, amazing

Hi Matt thanks for your videos. Love them. Lots of help. Question.... can you make a video explaining how to get good quality caricatures style from a selfish picture using midjourney? Thank you

good editing!

Good stuff!

Thanks for sharing, we had no idea you could go premium!

Been using this for a month! Thought I learned it from you? Oh geez. I should have filled you in. Ok. Next time. Your amazing Mr. WOLF 🐺

Yeah definitely insight face swap has been amazing

We need feeding updates and tutorials... Ai art Ai music Ai voice over Ai video Ai 3d models Great stuff matt ;)

You as Ken is killin' 😂 BTW all of the are really close!

@matt wolfe, have you ever tried to use "/describe" to get prompts from midjourney for your uploaded image? with this you should get results, that are looking more like the image you want to merge. If you then add additional prompts of new scenery and use the swopface tool in the end you may improve your final result. Let me know if this works. ;)

I feel like you're a little late to the party on this one Matt, this came out a few months ago! I found the best method is to use an image prompt and prompt to generate a photo and then using insight face - just like you showed in one of your methods. Was definitely the best output for me

Best AI content creator , hands down.

Tutorials and tool breakdowns all the way! 🤩

Thank you

I'd like to share this opinion with you. At the end of your video, you mentioned that you want to return to the type of videos you were producing on your channel. I wanted to let you know that, for me, I don't see a change per se. That is, to me, your channel signifies evolution. I believe that, as far as I'm concerned, the path and the variety of videos and topics you address on your channel doesn't warrant any changes. In other words, if you feel that your channel has changed, it's more because it has evolved. And that has happened because you're approaching it intelligently. Your decisions have led you to present this kind of content, which from my perspective is refreshing, always with your unique viewpoint that I value. It makes me feel like you're guiding me in this world of AI. Thank you so much for your contributions and congratulations.

Matt, thanks for your consistently awesome content. I'm searching for the best way to face swap in SD - on a Mac M1. I downloaded Roop successfully (SD confirms that I have it installed) via the extensions tab, but it doesn't actually show up in my txt2img or img2img generators. The only tutorials I've found on installation (quite a few) only give instructions for Windows - which involves also downloading Microsoft's Visual Studio. Am I to conclude Roop is not really available for Mac? No one actually says that. If that's true, do you know of a way to face swap that is compatible with my setup? Thx.

hello Matt. so is there a way to add your seeded characters to like a pic of 4 characters in a group

This was such a helpful video, thank you! I'm with you the whole way... except when I upgrade my image, I don't actually get the right click option (or the 3 dots) made available for the image. I can see them above in the InsightFaceSwamp bot messages, so I know I added it to my server, and I'm in the app... is it the starter plan I'm on? What could I be doing wrong?

I Made a video about this too , well done Matt 🎉

Hit the fkn goldmine on this page. Thank you Good Sir!

Another idea - you can upload a pic and have MJ give you 4 descriptions of that pic with /describe. And also use that as part of a prompt. Got that from another YT video.

I'm curious if you took one of the first 2 and told it to in swap again if it would keep getting closer to you. Just multiple iterations to keep getting closer. You're right about the face shape though, so it might not do much.

Your vids are great. I’m dying to turn my face into a simple cartoon in the style of Calvin and Hobbes.

Amazing Video 🤩🤩😻😻

Thanks!

hello, I really like your videos, but I saw that when you do the faceswap there is a "cut" and in the video you can hardly see that you rendered an image at 3:22 PM and that the faceswap ended at 3 :33 PM is there a wait of more than 10 minutes each time?

That image of Mona Lisa shown at 3.00 - was that created using Leonardo?

Gotta love some hat brows 😂!

8:40 lol 😂 You should tag Mr Musk 😂

great video. You made me laugh a lot ...

Awesome tutorial as always ... Question! I see that there are a limit of 50 credits on the basic plan... does anyone know if this is a daily limit or 50 free for lifetime credits?

Good Video

If you think about it they may not look like you but it's going off of the facial shape that's also defined by the shape of the body. If you were actually in whatever military or whatever then you might be surprised that those actually do look like you in those circumstances. Also they actually do look like you anyway even though they didn't change the hair and some things lol.

It works great. Bu the way do you kbow how to increase dayly credits for faceswap?thank you

Hello sir I can't see the three dot at the upper right side of my image that allows me to use the inswapper app. Neither can I even rightclick on the image when I give the prompt through my server. However when I go to direct messages i can see the three dot when i create images and even right click on the image. Please can u help?

I love the Nyan cat! Also, the video was very helpful as usual.

You are amazing💕💕💕💕💕💕

I'm having a hard time finding an answer for this online: Does upscaling on the newest versions of midjourney use any processing hours/credits?

Insightface is quite a name in the open source face recognition community. So I think (without having a look at their work in detail) they are tackling the problem from a different angle compared to like dream booth or other stuff. It's hard getting the model do create your face image cause you are unknown from training. You would have to manually figure out the semantic metrics the model is using to recreate a face looking similar to yours, which is a huge load of work and often requires changing of weights or huge back propagation efforts. This seems more like a non-learning approach. It's figuring out your face by face detection and the same with the face in the image and then merging these faces (like e.g. face morphing) until your identity is higher then the original identity, while preserving as much background information as possible. This leads to eyebrows on a hat. There is no semantic information for this, the semantics haven't been touched. Therefore you also cannot change semantics of the image without losing your identity face. But compared to the training process this is much faster. So I really like it. I'll have a look at their stuff later, thanks!

They all look close to you. Cheekbones and eyes are mainly used. And they are younger versions. You could make a photo album and claim it's you a decade ago or so.