Create Consistent Realistic Characters Using AI | Easy AI Image Tutorial

แชร์
ฝัง

ความคิดเห็น • 272

  • @SweetieNerdygirl
    @SweetieNerdygirl 7 วันที่ผ่านมา

    Best ai image generation tutorial !!! I successfully generated amazing photos without any problem !!!

  • @badashphilosophy9533
    @badashphilosophy9533 2 หลายเดือนก่อน +3

    i like how u showed the problem with examples to start the video, and i havnt watched the rest but i presume ur going to provide a good solution. u didnt waste time either, straight to the point which got u an instant like from me regardless of the rest of the vid just for your youtube presentation mastery lol. il subscribe if its practical and convenient enough for me to worry about

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      Glad you enjoyed the video!

  • @WORLDOFHEARTMUSIC
    @WORLDOFHEARTMUSIC 2 หลายเดือนก่อน +2

    "Big thanks for this beautiful and heartfelt work! In a time when AI dominates the digital landscape, thoughtful projects like this stand out. Let’s come together to support these authentic creations - with likes, comments, and subscriptions - so they reach the audience they deserve."

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      Appreciate you watching!

  • @Dr.X-Academy
    @Dr.X-Academy 25 วันที่ผ่านมา

    It’s very helpful. Thank you so much 💕☺️

  • @aashitajain3745
    @aashitajain3745 23 นาทีที่ผ่านมา

    Hello! I've seen some of your videos around face swaps and omg amazing work!!! I've been trying to find a way to swap the whole face, including the hair created using this method into a different pre-existing image (unlike taking one character and using text to image gen to create different contexts to place them in) and not been able to find a way to make it work. could you suggest a workflow or a tool that might work?

  • @LizKeye
    @LizKeye 2 หลายเดือนก่อน

    Still feel intimidated by comfy UI but I'm going to give this a go! Thanks so much, man.

  • @patrickzupanc1795
    @patrickzupanc1795 21 วันที่ผ่านมา

    Great video, thank you!

    • @curiousrefuge
      @curiousrefuge  16 วันที่ผ่านมา

      Glad you liked it!

  • @nobodyknew
    @nobodyknew 22 วันที่ผ่านมา

    Hahaha! I used to work with Jagger, funny to see her pop up randomly in this video.

  • @eKoush
    @eKoush 2 หลายเดือนก่อน +3

    oh yeah! this worked so well. Finally I can create unlimited pictures with the exact face of my cartoon character. This was so cheap, easy, and fast. omg. Now I can continue creating images with Midjourney and just add the suitable head on top. Or I can just generate them in the fal ai, since it looks really good, too.
    thank you very much!

    • @sarahashemi6521
      @sarahashemi6521 2 หลายเดือนก่อน +1

      Thank you for your feedback💖🙏. Unfortunately it doesn't free. Did you tested it for anime style characters? Im looking for a method to create anime characters with the same clothes. In many shots, the clothes must not be changed. Did you tested it?

    • @eKoush
      @eKoush 2 หลายเดือนก่อน

      @@sarahashemi6521 yes, it works well for anime style characters. only 10 pictures and it recreates it very well.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      Glad it helped you!

  • @davidiscove
    @davidiscove 2 หลายเดือนก่อน +2

    MASSIVE unlock. Thank you! Do you think the workflow could be expanded to include multiple consistent characters?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Thanks for watching!

  • @spiritualrealm555
    @spiritualrealm555 21 วันที่ผ่านมา +1

    How to train on two characters and use them together in one composition? Could you please make a tutorial on that too?

    • @curiousrefuge
      @curiousrefuge  16 วันที่ผ่านมา +1

      Thanks! We'll see what we can do!

  • @robertnewton9548
    @robertnewton9548 หลายเดือนก่อน

    This was a great video. You've sold me on your course. I'm curious if you've used Mystic (at Magnific) for upscaling vs Topaz.

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน +1

      We like Topaz for video upscaling

  • @XsweetstarliteX
    @XsweetstarliteX หลายเดือนก่อน

    This is great but can you do a video on which pricing option to choose in order to be able to do this? for novices, Fal is really confusing.

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      Unfortunately these pricing plans change very often so it would be outdated information fairly quickly. If fal is confusing, we'd suggest another good alternative which could be midjourney and kling.

  • @cielomaticaproducciones4747
    @cielomaticaproducciones4747 หลายเดือนก่อน

    Awesome video!
    How do you make the background for your talking head? It looks awesome!

  • @richardcharleshale
    @richardcharleshale 2 หลายเดือนก่อน

    Thank you so much, this was an epic video!!

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      You are so welcome!

  • @kelly_browning
    @kelly_browning หลายเดือนก่อน

    Awesome video! Just what I was looking for. Thank you for this!! How can we refine the images and add context similar to the thumbnails you showed of yourself in the beginning montage (closeup, with a lion, surfing, etc.)? Are there other programs you used in addition to this for that workflow?

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      Glad it was helpful! We use Akool to help with thumbnails

    • @kelly_browning
      @kelly_browning หลายเดือนก่อน

      Thank you so much! I'm about to purchase GigaPixel based on your recommendation but I want to be able to support your channel. This video was a HUUUGE lifesaver! Do you have an affiliate link I can use so that you'll get credit?

  • @Ancestor_Numero_Uno
    @Ancestor_Numero_Uno 20 วันที่ผ่านมา

    Nice tut! You think the files can be used in InvokeAI?

    • @curiousrefuge
      @curiousrefuge  16 วันที่ผ่านมา +1

      No idea! We'd need ot test!

  • @Eliminatorxtreme
    @Eliminatorxtreme 2 หลายเดือนก่อน

    Thanks for the swagger. Now I've got the moves like Jagger. 😂

    • @Eliminatorxtreme
      @Eliminatorxtreme 2 หลายเดือนก่อน

      I'll see my way out.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      hahaha we appreciate the jokes ;)

  • @ChadNeo
    @ChadNeo 2 หลายเดือนก่อน

    Great video 👌🏾👌🏾👌🏾

  • @amicrazy195
    @amicrazy195 2 หลายเดือนก่อน

    Great tutorial

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Thank you! Cheers!

  • @faizalbulkini
    @faizalbulkini หลายเดือนก่อน

    thank you so much

  • @jdcmsigma47
    @jdcmsigma47 หลายเดือนก่อน

    whoa you are good!!

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      thanks for watching!

  • @Jerell_AI
    @Jerell_AI 2 หลายเดือนก่อน

    Helpful video!

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Glad you think so!

  • @heartshinemusic
    @heartshinemusic 2 หลายเดือนก่อน +10

    I thought that Midjourney distorts real uploaded humans on purpose. They suggest creating a custom character on their platform and use that as a character reference.

    • @EdoardoScognamiglio
      @EdoardoScognamiglio 2 หลายเดือนก่อน

      What you mean with “custom character”?

    • @heartshinemusic
      @heartshinemusic 2 หลายเดือนก่อน

      @@EdoardoScognamiglio I meant designing a character that you later want to use as a consistent character. But creating it on MJ from scratch and not starting from an existing photo.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We haven't heard anything about that

    • @heartshinemusic
      @heartshinemusic 2 หลายเดือนก่อน

      @@curiousrefuge I could swear I read something like that on the MJ user manual page when consistent characters were introduced.

    • @artvsmachine
      @artvsmachine 2 หลายเดือนก่อน +2

      @@curiousrefuge It's in their documentation. They say they will deliberately distort any celebrity, but they do it to unknowns as well.

  • @PaulFishwick
    @PaulFishwick 2 หลายเดือนก่อน +4

    Jagger is in a coffee shop. I agree that there is a good consistency for Jagger. What about the coffee shop? They are all different . Is there a way to keep the scene (coffee shop interior) consistent as well?

    • @adrianmunevar654
      @adrianmunevar654 2 หลายเดือนก่อน +4

      You would need a LoRA for every consistent aspect you need: character, location, clothes, etc. Or becoming a master in controlnet or canny.

    • @HarryClipzFilmz
      @HarryClipzFilmz 2 หลายเดือนก่อน

      @@adrianmunevar654So create a character, create a environment, then train them both together or what can we do? I want the same thing

    • @adrianmunevar654
      @adrianmunevar654 2 หลายเดือนก่อน

      @HarryClipzFilmz it is more effective to train them separately. LoRA for the character with the clothes, and LoRA for the location. Why? Because before you start training you have to caption every image. In the character's case you gotta write the name and gender in every caption, then the description of the clothes and the different expressions, camera angles, etc. So if besides it, you gotta describe the location, it would end up being a mess, and the results would be not the best. So, if you train just the character first, and then the environment, then you can train more environments and put the character in different locations just by including the character LoRA + the desired environment LoRA, and if you want, a style LoRA as well. It sounds tricky but in practice it is easier.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Great point! In order to get consistency with the location/environment, it would require lots of inpainting or training on a location

  • @justinlloyd3
    @justinlloyd3 2 หลายเดือนก่อน +1

    Thank you!

  • @ViralBlog101
    @ViralBlog101 2 หลายเดือนก่อน

    This is very cool. I am doing experiments with the same 💢🤩

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Glad you think so!

  • @salomeshunamon
    @salomeshunamon หลายเดือนก่อน +1

    Any tips for consistent clothing?

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      We suggest using tools like Pika, training a lora, or even Kling

  • @garrydye2394
    @garrydye2394 2 หลายเดือนก่อน +4

    Question about your courses. By the time you put in all the effort to make a course...isn''t it fairly outdated?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +3

      We update the courses on a monthly cadence

  • @FANTASTICWONDERFUL
    @FANTASTICWONDERFUL 2 หลายเดือนก่อน +4

    🔥Nice. Is there a way you can get that level of character consistency with characters you AI generate?

    • @aaronhhill
      @aaronhhill 2 หลายเดือนก่อน

      This is my question as well.

    • @juanestebanmarin9244
      @juanestebanmarin9244 2 หลายเดือนก่อน +5

      Yes, in midjourney you can expand your photo to 16:9 and move it to the left side grid then u promt "a two panel grid of photos of "your character description", diferent angles " and midjouney Will give you ur character in diferent poses position and angles

    • @aaronhhill
      @aaronhhill 2 หลายเดือนก่อน +1

      @@juanestebanmarin9244 Perfect! Thanks for this! 😊

    • @FANTASTICWONDERFUL
      @FANTASTICWONDERFUL 2 หลายเดือนก่อน

      @@juanestebanmarin9244 Thank you, I will try that

    • @armondtanz
      @armondtanz 2 หลายเดือนก่อน +1

      ​@@FANTASTICWONDERFUL
      Not 100% . I'm literally doing it today and they coming out very different.
      Or not 100% anyway. 😢

  • @justbenx
    @justbenx 2 หลายเดือนก่อน

    I’d be curious to see this done on animated or cartoon characters as well.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      It does well on animated characters but definitely worth testing!

  • @sarahashemi6521
    @sarahashemi6521 2 หลายเดือนก่อน +1

    Thank you. I need a method to create anime style characters. Also, i need to keep the clothes the same in many frames. Did you test it? Since it is not a free method, i don't have any way to test it.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We didn't test it but others in the comments have said it works well with that style

  • @videosfeoscomotucara9038
    @videosfeoscomotucara9038 2 หลายเดือนก่อน +4

    More like "consistent character" plissss

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We appreciate you watching!

  • @lineardrift
    @lineardrift 2 หลายเดือนก่อน

    Thank you so much!!! This was very helpful. The issue is when you have more than one person in the scene. I have my lead character in a group of 4 people having coffee. How do I put her face on the right body? LOL I am working on part 2 of my film and this video was very helpful. Any thoughts on multiple people in one shot?

    • @armondtanz
      @armondtanz 2 หลายเดือนก่อน +1

      In freepik you can inpaint the person using an image reference. You can do this in MJ as well.
      Not 100%. I have to tweak in photoshop afterwards.

    • @lineardrift
      @lineardrift 2 หลายเดือนก่อน

      @@armondtanz Let me try freepik. Thank you!

    • @armondtanz
      @armondtanz 2 หลายเดือนก่อน +1

      @@lineardrift wait. I'm not sure u can in freepik.
      Its defo in midjourney.

    • @armondtanz
      @armondtanz 2 หลายเดือนก่อน

      @@lineardrift theres a great tutor called gilbatree,
      His channel has great hacks on how to use MJ to get best face fix results.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      Many people will use the inpainting trick with a reference photo to get a second character

  • @gospelsongsoversins
    @gospelsongsoversins 2 หลายเดือนก่อน +2

    Thanks a lot for sharing but what about consisten environment or background scenes, lets say if its a train station how can i make sure the character background, the room, etc etc stay consistent

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Train different LORAs or you can also do lots of in painting, or lastly use a 3D environment so it stays consistent

  • @jonathanrice2568
    @jonathanrice2568 2 หลายเดือนก่อน

    Great thanks. By the way - HOW do you get such a clean background swap please?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      You can do it in Runway or any video editor too!

  • @RomainGal
    @RomainGal 2 หลายเดือนก่อน +2

    Why upload many images if you click on inference button just for one specific image ? Why is not a merge of everything ?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Not sure I understand what you mean

  • @wonderlandfortina
    @wonderlandfortina 2 หลายเดือนก่อน

    Thank you for your video. I wonder for the upgrade of the image quality, if I already have Topaz AI video, do I still need to buy Gigapixel again?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      Yes they are different. while there are many free tools, Topaz AI Video (for video) and Gigapexiel (for photos)

    • @wonderlandfortina
      @wonderlandfortina 2 หลายเดือนก่อน

      @@curiousrefuge thank you for your reply.

  • @mili6785
    @mili6785 หลายเดือนก่อน

    hello, Thank you for the video, I have a question: can lora control the cloth on the character in different scene

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      Yes! Many times people train multiple loras. The easiest route would be to use Kling or Pika in this case though.

    • @mili6785
      @mili6785 หลายเดือนก่อน

      @@curiousrefuge OK, Thank you for the answer

  • @sarahashemi6521
    @sarahashemi6521 2 หลายเดือนก่อน

    Thank you again. Could you please test it for anime style characters and try to keep the clothes? Before paying for this platform, I'm looking for a way to be sure it is a correct choice 💖🙏❤️

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      I can try

    • @sarahashemi6521
      @sarahashemi6521 2 หลายเดือนก่อน

      @@curiousrefuge thank you very much ❤️

  • @roshdialkadri328
    @roshdialkadri328 2 หลายเดือนก่อน

    That's cool maybe for magazine images...but consistency from shot to shot in a narrative piece is the problem where image to video is implemented, that is keeping the exact same look, wardrobe, hair, etc. in a scene

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      It definitely has some work to be consistent across tons of different scense (especially for the environment)

  • @Gheon_Stonecamp
    @Gheon_Stonecamp 2 หลายเดือนก่อน +3

    So what Id like to know is. Is this better than Krea's Lora training which also uses Flux?

    • @ecanus-3605
      @ecanus-3605 2 หลายเดือนก่อน

      No, from what I can tell it's "simpler" but by now it is easy enough to setup training on fe. Runpod, best option is still comfy.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We'll have to do some tests!

    • @Gheon_Stonecamp
      @Gheon_Stonecamp 2 หลายเดือนก่อน

      @@curiousrefuge Yeah, Krea is pretty good with Loras also, but i would also need to compare and test. Thank you for the tips on this, though. Dig this channel. I learn a lot.

    • @joshxt
      @joshxt หลายเดือนก่อน

      @@Gheon_Stonecamp any updates? Krea is definitely more straightforward

    • @Gheon_Stonecamp
      @Gheon_Stonecamp หลายเดือนก่อน

      @@joshxt Way more straight forward. Krea is actually a really good tool to have. They are kinda like the swish army knife of the AI Art world I think.

  •  2 หลายเดือนก่อน

    Thanks for the platform comparison. How would you rate Krea’s character training with Flux?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We'll do some testing!

  • @PIXALRIOTS
    @PIXALRIOTS หลายเดือนก่อน

    Can you use the llore model in mid journey?

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน +1

      No, midjourney uses it's own techniques for consistency.

    • @iKINDREAD
      @iKINDREAD หลายเดือนก่อน

      @@curiousrefugethat’s too bad. Wouldn’t be cool if you can build a lore and use in MJ

  • @davidevichi6484
    @davidevichi6484 2 หลายเดือนก่อน

    Wow! Can't wait to try this! Is there a way to do style reference (similar Midjourney) with Flux1.1 pro? Thanks! ;-)

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      Anything is possible in comfy but we'll look and see what easier options are out there!

  • @planetofshireen
    @planetofshireen หลายเดือนก่อน +1

    Hi , I wanted to ask if we can edit image in FLUX (outpaint/inpaint) as it is available in Midjourney.

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      In ComfyUI you can

  • @AliciaBankhoferEdu
    @AliciaBankhoferEdu 2 หลายเดือนก่อน

    My question is does it HAVE to have one or more reference images?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      The more references images to train the Lora the better

  • @eliezerborges
    @eliezerborges หลายเดือนก่อน

    I spent 10 bucks on it and I could generate, let's say... 20 images! I think I'm doing something wrong. There's a way to spend less? When you buy $10 on credit they say:
    Approx. 4.5k images using SDXL with defaults.
    4.5k images? I could generate only 20!!! damnnnn lol
    Thanks for the nice video!

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      Hope in our discord and we can try and troubleshoot a better strategy for ya!

    • @eliezerborges
      @eliezerborges 28 วันที่ผ่านมา

      @@curiousrefuge Can you send me the link? Thanks!

  • @High-Tech-Geek
    @High-Tech-Geek 2 หลายเดือนก่อน +1

    Q: Is Gigapixel good for upscaling illustrations, anime, sketches or non-photorealistic images?

  • @umairsadique
    @umairsadique 2 หลายเดือนก่อน

    I want to create images of my clone... which one will be best for long term please guide.....

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Please watch the video!

  • @scottdunlap4109
    @scottdunlap4109 หลายเดือนก่อน

    Im just getting started, not even a novice. Is there something comparable that's free?

    • @curiousrefuge
      @curiousrefuge  25 วันที่ผ่านมา

      For free, feel free to check out our free course at our website (and of course youtube has tons of great material!)

  • @chelseymark5070
    @chelseymark5070 2 หลายเดือนก่อน

    Thank you!!!!!!!! This is so much better for consistent characters!!!!!

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Glad you like it!

  • @gerardomartinez9486
    @gerardomartinez9486 26 วันที่ผ่านมา

    Hola CR. Quiero usar IA para generar modelos para hacer arte, que app me recomiendas ? Quiero combinar dibujos y fotografía para obtener modelos para pintar. Help. Saludos

  • @MalliMelloStudio
    @MalliMelloStudio 2 หลายเดือนก่อน

    I've been looking for a way to create illustrated characters consistently for a long time. I have created lots of monsters in a handdrawing style. The style is consistent but I only have one picture of each monster and I just can't find a way to duplicate the monsters in that style. With illustrations, the characters don't have to be 100% consistent. I trained a lot of datasets. Maybe I'm just not capable... :D

    • @ecanus-3605
      @ecanus-3605 2 หลายเดือนก่อน

      Have you tried using "Instant Lora" by AloeVera and Img2img to get your dataset started?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      You trained multiple datasets of this character and it didn't work?

  • @Samt2b
    @Samt2b 2 หลายเดือนก่อน

    Can you make multiple consistance

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Yes, it often requires some kind of inpainting though.

  • @emeth6322
    @emeth6322 หลายเดือนก่อน

    What image quality you used ?

    • @emeth6322
      @emeth6322 หลายเดือนก่อน

      300$ phone images are enough?

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      yes

  • @ArtificialGeekStudio
    @ArtificialGeekStudio 2 หลายเดือนก่อน

    What if i use magnific to upscale and keep resemblance at max?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We'll have to test!

  • @Sanjaypatelxm
    @Sanjaypatelxm 2 หลายเดือนก่อน

    i trained two characters individually, how can i use this both in one image generation them together. please make video. i am trying from many days but not successful, ai makes both characters mixed.😢😢😢😢

    • @trumseo
      @trumseo 2 หลายเดือนก่อน

      you can use shaker ai to train 2 characters, then use them in the same image

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We'll have to do some testing.

  • @RAMAMOORTHI.A
    @RAMAMOORTHI.A 2 หลายเดือนก่อน

    I want to crate comic style 2d images is FAL AI is best?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      It will still work for anime/cartoonas

  • @mohammadhussain4547
    @mohammadhussain4547 2 หลายเดือนก่อน

    I get plastic kind of effect after training LORA. I have tried training LORA couple of times on replicate, the option you mentioned above etc but all gives that plastic kind of feel. I usd one platform where I was able to generate through Cyber Realism model more like a photograph or realistic image as someone captured it through Camera and did post processing. Do you have any idea if somehow I can achieve those results?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      That’s definitely a common issue, it might be due to the dataset used. Where did you download your lora

    • @mohammadhussain4547
      @mohammadhussain4547 2 หลายเดือนก่อน

      @@curiousrefuge Sorry couldn't get you.
      I downloaded LORA from my huggingface training to my Local drive.

  • @PanayiotisMavrokefalos
    @PanayiotisMavrokefalos 25 วันที่ผ่านมา

    Cam the subject be an item? Or does it have to be a human subject?

  • @ditapeskova4
    @ditapeskova4 2 หลายเดือนก่อน

    Have you also tried to train it to keep the clothes the same? As that's more of an issue when it comes to creating consistent images. Prompting is inadequate for that.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We'll give it a shot!

  • @rocren6246
    @rocren6246 24 วันที่ผ่านมา

    what about sd loras? Flux is too limited.

    • @curiousrefuge
      @curiousrefuge  23 วันที่ผ่านมา

      We can certainly test!

  • @nickfoister
    @nickfoister 2 หลายเดือนก่อน +1

    How much does it cost?

    • @schultzeworks
      @schultzeworks 2 หลายเดือนก่อน +2

      This question seems so basic and fundamental, it shocks me that it was omitted.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      check the website to see if there is a sale!

  • @mertyigit4447
    @mertyigit4447 หลายเดือนก่อน

    How can I convert a prompt into a template. I want to make @Jagger look like spiderman-3 and I want to same output on @Matt look like spiderman-3

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      GPT4 paid lets you create your own custom GPTs where you can configure it to be prompts based on what youa re looking for.

  • @Marvelgirl_1682
    @Marvelgirl_1682 2 หลายเดือนก่อน

    How can I create a video for my Virtual model

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      You can use Runway!

    • @Marvelgirl_1682
      @Marvelgirl_1682 2 หลายเดือนก่อน

      @ thanks for the response. Do we have any other free tool for making videos? Please suggest

  • @ldevs430
    @ldevs430 2 หลายเดือนก่อน

    I've been doing it for months but had to use external apps

  • @mohanapatki
    @mohanapatki 3 วันที่ผ่านมา

    Does anyone use Runway's Image Upscaler? How does it hold up against Topaz and others like it?

  • @EvolvingEdens
    @EvolvingEdens 2 หลายเดือนก่อน

    how to put 2 consistent character in the same image

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      there are different ways but they typically involve inpainting a second character whether it's through MJ or a Lora

  • @haljordan1575
    @haljordan1575 2 หลายเดือนก่อน +1

    It can't be consistent characters if the clothes are not.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      You can train a clothing model using this technique. Just make sure the training data has the same clothes.

    • @Sanjaypatelxm
      @Sanjaypatelxm 2 หลายเดือนก่อน

      ​@@curiousrefugehow to train clothing model

    • @sarahashemi6521
      @sarahashemi6521 2 หลายเดือนก่อน

      Im looking for the same clothes too. In a real movie in many frames, the character must have the same clothes.

    • @haljordan1575
      @haljordan1575 2 หลายเดือนก่อน

      @@sarahashemi6521 these guys are always hyping things that don't work as well as they should.

  • @Beretta0072
    @Beretta0072 2 หลายเดือนก่อน +1

    im desperatly looking for a tutorial that will help me with a pixar charcter consistant character. Im have just one or two consistant images at present. Without using a real life person its hard to get waht i need to train a lora. There are lots of tutorials around but they are all using real people. Can anyone help. Im trying to create a book for my niece who was born deaf and with learning difficulties, she only 2 and would make the best christmas present

    • @RoganGriffin
      @RoganGriffin 2 หลายเดือนก่อน

      I found a very useful way around that. Throw it at a video generator like Runway. You can ask it to do turnaround views, or give it a reference of lots of different expressions using Act One. Then you can pull out individual frames and use those to train on.

    • @Beretta0072
      @Beretta0072 2 หลายเดือนก่อน

      I've actually done the same. With the character I have. Would be interested in sharing my material and getting some advice.

    • @adrianmunevar654
      @adrianmunevar654 2 หลายเดือนก่อน +1

      You can ask Flux in ComfyUI or another UI to generate a character sheet in several poses and expressions, after that you train the LoRA and you can generate endless images. Or, if you want to, I could make it for you

    • @Beretta0072
      @Beretta0072 2 หลายเดือนก่อน

      @@adrianmunevar654 - id really appriciate that. I have comfy but the templates i used didnt work well, can i share what i have with you soemwhere, perhaps in curious refuge discord

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Try using it a reference image in MJ, then for your prompt ask for a multi directional character sheet.

  • @Romeo615Videos
    @Romeo615Videos 2 หลายเดือนก่อน +1

    what is this the free way to do stuff or still on a pay wall? do you have a discord?

    • @ecanus-3605
      @ecanus-3605 2 หลายเดือนก่อน

      You can train for "free" on civit (needs Buzz but that can be collected for free)

    • @Romeo615Videos
      @Romeo615Videos 2 หลายเดือนก่อน +1

      What does that even mean?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      We do have a discord, please check out our website and learn mroe !

    • @Romeo615Videos
      @Romeo615Videos 2 หลายเดือนก่อน +1

      @@curiousrefuge Nice site but im not looking to buy a course I have won awards for my film work im just not up on building the tools yet

    • @07cali4nia
      @07cali4nia 2 หลายเดือนก่อน

      @@Romeo615Videossame here. Also not interested in paying $729/yearly until I cancel either… that’s a bit much.

  • @ecanus-3605
    @ecanus-3605 2 หลายเดือนก่อน

    I wouldn't call using Comfy advanced... also there is "instant Lora" by AloeVera which makes consistent characters from 1 img to multiple to lora quie easy.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Perhaps it's all relative! Comfy can get quite advanced depending on the workflow...

  • @adarwinterdror7245
    @adarwinterdror7245 2 หลายเดือนก่อน +7

    The big question is how do I create consistent characters from scratch using AI.
    Its possible with MJ, but THATS the big problem. If i could have enough images of the same character to train the LORA than it means I'm already past the problem 😄

    • @RoganGriffin
      @RoganGriffin 2 หลายเดือนก่อน +6

      I found a very useful way around that. Throw it at a video generator like Runway. You can ask it to do turnaround views, or give it a reference of lots of different expressions using Act One.

    • @Larimuss
      @Larimuss 2 หลายเดือนก่อน

      You need to use comfyui or some local/cloud run private machine and load the character workflows etc. It's the only way currently. Which gives you character sheet, but a bunch of different portrait emotions, and you can also gen other images in the same single run. Only way essentially is many images out of one generation. By utilising the same seed, ksampler, latent, etc in a way to produce many photos of the same character to train.

    • @adarwinterdror7245
      @adarwinterdror7245 2 หลายเดือนก่อน

      @@RoganGriffin Dude!! That is a very good idea!
      Not perfect, but as you said - a workaround.
      That's the kind of solution I would have expected myself to come up with, but didn't.
      very clever. I really appreciate the tip.
      Such a smart idea!

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      Many people will generate multiple angles of the same character. And also the Runway trick maight work!

  • @MuslimKidStories-2025
    @MuslimKidStories-2025 2 หลายเดือนก่อน

    is it free or paid. can u tell me

  • @idabrinck-lund9744
    @idabrinck-lund9744 หลายเดือนก่อน

    Cant make it work - cant find support - feel so lost with ALL these opportunities and so poor support and SOOO many providers

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      Join our discord and let us know howe can help!

  • @divergentlife493
    @divergentlife493 หลายเดือนก่อน +1

    Thumbs down! Makes completely and utterly no sense. Says to select lora fast training. No clue what that means. No information on what model section is for. I have an image of a male character and another of a female character. Color artwork not photo. I simply need both characters holding hands running away from danger.

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      It sounds like you're looking for something a little different!

  • @artvsmachine
    @artvsmachine หลายเดือนก่อน

    Good, but the quality is too low, and it's also rather expensive.

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      Each month it gets better and better : )

    • @artvsmachine
      @artvsmachine หลายเดือนก่อน

      @@curiousrefuge Does it? Wish they'd do 1080X1920. It's the standard for video, sooooo, why not? Meanwhile, if I've already paid to create the model, why am I paying more for each generation? I dig it. I use it. But if it could just go that bit more. Incidentally, it outperforms MJ at CREF, so scaled back my MJ membership. But, you know what's really cool. Vidu can use 1-3 references and go right to video and does a great job. Seem more accurate the FLUX.

    • @artvsmachine
      @artvsmachine หลายเดือนก่อน

      @@curiousrefuge Leonardo's version does higher rez with the option to upscale. Not sure if it's as versatile yet, though. Gotta test it against Krea as well, which allows multiple uploads. The Flux version is too small and too costly considering.

  • @The.HandbagHaven
    @The.HandbagHaven หลายเดือนก่อน

    Is this free or paid?

  • @RAMAMOORTHI.A
    @RAMAMOORTHI.A 2 หลายเดือนก่อน

    one basic question - Is fal ai is website which offers all ai image and video creation tools in one place with the required cloud system specs? if i signup FAL AI i can create consistent Images?

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      It doesn't do everything. We simply suggest it for photos

  • @cinematic_monkey
    @cinematic_monkey 2 หลายเดือนก่อน

    Zoom in on the eyes, would ya. Lovely detail there. Gotta love the reptillian vibe going on there.

  • @theaieyee
    @theaieyee 2 หลายเดือนก่อน

    Is it free?

  • @SufyanBloch-vy7ps
    @SufyanBloch-vy7ps 28 วันที่ผ่านมา

    Is this free?

    • @curiousrefuge
      @curiousrefuge  25 วันที่ผ่านมา

      Afraid most of these tools are not!

  • @janvanlaarhoven1125
    @janvanlaarhoven1125 2 หลายเดือนก่อน

    why not use photo ai? its much better

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      We'll have to test...

  • @Azzalto40
    @Azzalto40 2 หลายเดือนก่อน +2

    better than Midjourney you kidding? 🤣🤣🤣🤣

    • @buster5661
      @buster5661 2 หลายเดือนก่อน +2

      when you have to pay for something, its usually better

    • @WAITERS.CHENTHU
      @WAITERS.CHENTHU 2 หลายเดือนก่อน +2

      In terms of character consistency yes. Better than MJ

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +3

      For consistency of a character. Did you see the tests?

    • @Azzalto40
      @Azzalto40 2 หลายเดือนก่อน

      @@curiousrefuge in mid journey?? if you use --cref + cw 100 will work

  • @buster5661
    @buster5661 2 หลายเดือนก่อน

    i would much prefer to use ai that i can use offline

    • @AI3EFX
      @AI3EFX 2 หลายเดือนก่อน

      Open source is your friend 💻

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      If you have a good computer you can! :)

    • @AI3EFX
      @AI3EFX 2 หลายเดือนก่อน

      @@curiousrefuge Fax 📠
      Preferably with a competent GPU

    • @buster5661
      @buster5661 2 หลายเดือนก่อน

      @@curiousrefuge surprised Pikachu face

  • @lucifer9814
    @lucifer9814 29 วันที่ผ่านมา

    Another click bait, I knew this was another one more of those promotional crap right off the bat. Next time rephrase your titles better, "Create consistent Realistic Characters with lora" would be a far more ideal title for this video.

    • @curiousrefuge
      @curiousrefuge  28 วันที่ผ่านมา

      Sorry you felt this way!

  • @MrJvog
    @MrJvog 21 วันที่ผ่านมา

    you forgot to mention it costs money for every generation, just skipped riiiiight past that part. massive dislike and don't recommend.

    • @curiousrefuge
      @curiousrefuge  16 วันที่ผ่านมา

      Please assume that any tool is paid.

  • @AdrianSchoeman
    @AdrianSchoeman 4 วันที่ผ่านมา

    Fantastic tutorial - thanks! For consistent realistic characters in video, Kling's Custom AI model works so well! Have a look at my short first project: th-cam.com/video/jkuZOIgReqI/w-d-xo.html

  • @ShifterBo1
    @ShifterBo1 2 หลายเดือนก่อน

    Its paid, need 2 dollars to train the model

  • @AILab0001
    @AILab0001 2 หลายเดือนก่อน

    🤨better than Midjourny? 😂Leonardo AI does a better job and it’s even easier to use

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      To each their own!

  • @paul.hnewman9784
    @paul.hnewman9784 2 หลายเดือนก่อน

    Freepik is better for this.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      We'll do some testing!

  • @ssaeedhy
    @ssaeedhy 16 วันที่ผ่านมา

    spoiler alert! its not free

    • @curiousrefuge
      @curiousrefuge  15 วันที่ผ่านมา

      Spoiler! :) Nearly any tool isn't free aside from ComfyUI run locally! :)

  • @justimagine.prod_ai
    @justimagine.prod_ai 2 หลายเดือนก่อน

    Krea is much better

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน

      Krea is definitely a cool contender!

  • @AdamFit-b6w
    @AdamFit-b6w หลายเดือนก่อน

    Not free wasting time

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน

      There is no such thing as a free tool (whther it's paid money or your data as the cost).

  • @behzad4x
    @behzad4x หลายเดือนก่อน +14

    This model is not free. But the video creator doesn't say a single word about it and teaches as if it were free!

    • @curiousrefuge
      @curiousrefuge  หลายเดือนก่อน +7

      Please consider almost all AI models not being free.

    • @PaintIt4Ward
      @PaintIt4Ward 25 วันที่ผ่านมา

      Get a job lol

    • @bprzy852
      @bprzy852 8 วันที่ผ่านมา

      Do you realize how expensive the compute is for most AI software to run? Anything you’re not able to run on your own machine is cost somebody something. Which means even if those things are free you’re paying in some other fashion. The best AI results with heavy compute are all going to cost something.

  • @sqvaifilmmaking
    @sqvaifilmmaking 2 หลายเดือนก่อน

    Hi, I've found interesting the site in replicate /fofr/consistent-character, to generate different poses from only one image of your character before creating its lora.

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      thanks we'll check it out!

  • @stunninghollywood
    @stunninghollywood 2 หลายเดือนก่อน

    Thanks for all the knowledge you share to help us complete our film, first teaser at stunninghollywood

    • @curiousrefuge
      @curiousrefuge  2 หลายเดือนก่อน +1

      thanks we'll see how it looks!

    • @stunninghollywood
      @stunninghollywood 2 หลายเดือนก่อน

      @@curiousrefuge Great of you. Extended first look will be uploaded today.