Create Consistent Characters for Ai Stories with FLUX!

Поділитися
Вставка
  • Опубліковано 27 лис 2024

КОМЕНТАРІ • 147

  • @Natureoai
    @Natureoai Місяць тому +13

    First view ... ♥ pin please 🥰

  • @raztube90
    @raztube90 Місяць тому +12

    I tried this with mid journey and it worked ….
    Describe your character in details (use chat gpt)
    Put this at the beginning
    A character sheet, realistic photography, simple background, multiple views, from multiple angles, visible face, portrait ,neutral expression, photography of (your characters description)
    Downlaod and Upsample the best pic and crop them
    Upload to mid journey and press character icons on all of them
    Prompt and creat (no need to describe what she is wearing everytime) I saw that mid journey worked perfectly everytime 🎉

    • @arniifitness
      @arniifitness Місяць тому +1

      Thanks for sharing it worked with midjourney! was literally my first thought watching this

    • @turbanbasscity272
      @turbanbasscity272 Місяць тому +1

      Is "upsample" same as upscaling?

    • @hamloji
      @hamloji Місяць тому

      Yes​@@turbanbasscity272

    • @taoprompts
      @taoprompts  Місяць тому +2

      Midjourney works pretty well for human characters, although it struggles much more with nonhumanoid characters

    • @Razkoh
      @Razkoh 25 днів тому +1

      ​@@taoprompts is niji the best for anime or manga consistent characters?

  • @Pusty159
    @Pusty159 3 дні тому

    Thank you ! You made one of the most useful guides. You gave me a very valuable tip that solved my problem when creating a video. Greetings from Poland

  • @PokeGamer-jz2fe
    @PokeGamer-jz2fe Місяць тому +7

    Thanks love from India 🇮🇳

    • @taoprompts
      @taoprompts  Місяць тому +1

      Great to hear from India! I appreciate the love

  • @YouUsAndAI
    @YouUsAndAI Місяць тому

    Amazing - these new developments are so exciting!! Thank you so much for the step by step instructions Tao!!

    • @taoprompts
      @taoprompts  29 днів тому +1

      Yeah, it's pretty amazing the see how consistent the Ai can be with these characters 👍

    • @YouUsAndAI
      @YouUsAndAI 28 днів тому

      @@taoprompts So true!

  • @MicDoesThings
    @MicDoesThings Місяць тому +2

    Love your contents! Always on top of the latest trends.

    • @taoprompts
      @taoprompts  Місяць тому

      Thanks 👍! Tuning Flux models seems to be the meta now for characters

  • @Juninhokei
    @Juninhokei Місяць тому +1

    I bought Openart and I'm very surprised whith high quality ❤❤

    • @taoprompts
      @taoprompts  Місяць тому

      Yeah, it's also got so many different features to try 👍

  • @giri.goyo_yt
    @giri.goyo_yt Місяць тому

    Excellent work, once again, Tao! Can't wait to try this! 🎥

    • @taoprompts
      @taoprompts  29 днів тому +1

      Thanks Giri! Finetuning flux models looks like the meta right now for consistent characters

  • @MysticDepths-24Hrs
    @MysticDepths-24Hrs Місяць тому +4

    Great Tutorial Bro

    • @taoprompts
      @taoprompts  Місяць тому

      Thanks man! Its fun to try new ai tools

  • @jasonbraithwaite9204
    @jasonbraithwaite9204 14 днів тому

    Excellent video, works perfectly, thank you

    • @taoprompts
      @taoprompts  11 днів тому

      Great to know you got it working!

  • @YouUsAndAI
    @YouUsAndAI 24 дні тому

    So brilliant, Tao, I just tried it out! Thank you Brother!

    • @taoprompts
      @taoprompts  19 днів тому +1

      Great to hear you gave this a shot! I'll try to make more guides with characters like this

    • @YouUsAndAI
      @YouUsAndAI 19 днів тому

      @@taoprompts That would be great! Thank you!

  • @thewebstylist
    @thewebstylist Місяць тому +1

    Jumpin on this tomm Tao bro! 🎉

    • @taoprompts
      @taoprompts  Місяць тому

      I appreciate you man. I think tuning Flux models will be the meta for a while

  • @naidecualquiera663
    @naidecualquiera663 29 днів тому

    Thanks for all the steps!

  • @redbunnyclassic
    @redbunnyclassic Місяць тому

    Another great video!
    Thank you 👍🏽

    • @taoprompts
      @taoprompts  29 днів тому

      Thanks Bunny! Tuning flux models seems to work best right now for consistent characters

  • @tobypointer
    @tobypointer Місяць тому

    Excellent I've been trying to figure this out, Thanks

    • @taoprompts
      @taoprompts  29 днів тому

      Thanks Toby! Midjourney really needs a feature like this, right now finetuning flux models seems to be the best way to create characters

  • @mehekkhankhan2081
    @mehekkhankhan2081 Місяць тому

    Hey Tao Prompts!
    First off, thanks so much for the amazing content you share! I watched your video on creating dynamic poses, which was super helpful. Now, I'm hoping you could do a more detailed tutorial on how to create dynamic fighting poses for comics using AI, while still keeping each character's unique traits intact.
    I'd love to see some tips on making these poses as realistic and intense as possible - ideally with free tools, if that's doable. It would be a huge help for bringing action scenes in my comics to life!
    Thanks again for everything you share. Looking forward to learning more!

    • @taoprompts
      @taoprompts  Місяць тому

      Thanks for the suggestion! You will need some kind of ControlNet for pose control

    • @mehekkhankhan2081
      @mehekkhankhan2081 Місяць тому

      ​@@taoprompts
      Thanks for getting back to me! I’d really appreciate it if you could make a video on creating dynamic fighting scenes using AI. Seeing a step-by-step video tutorial would help a lot with bringing those intense action poses to life in my comic.
      Thanks again for considering it, and looking forward to learning more from you!

    • @michaelyeboah7129
      @michaelyeboah7129 29 днів тому

      @@taoprompts which video explains how to make dynamic poses?

    • @taoprompts
      @taoprompts  29 днів тому

      @@michaelyeboah7129 this one for leonardo AI ua-cam.com/video/rIsus4VLVTw/v-deo.html

  • @ainashaubakirova2256
    @ainashaubakirova2256 26 днів тому

    Thanks a lot Tao, great review as always!
    I'm curious regarding a couple of things and I'd like to hear your opinion about them...
    1. Flux model training you explained - can it be done in Replicate (your earlier suggestion)? Or Open Art is more functional?
    2. Do you think it can be possible to train a character - yourself same way?
    3. Do you have any solution for issue: I've got a character I want and desired background is another image... What would be the best way to merge them? If you have that experience...

    • @taoprompts
      @taoprompts  19 днів тому

      Yes you can absolutely train yourself with this.
      Flux finetuning is pretty much the same on replicate or this site, I'd say replicate is more customizable but requires a bit more background knowledge.
      To add in a specific background and make it look good, you may want to try photoshop

  • @MettameowChannel
    @MettameowChannel Місяць тому

    I just bought unlimited plan because of you thanks again for sharing your awesome knowledge🎉

    • @taoprompts
      @taoprompts  Місяць тому

      That's awesome! Openart has so many different features to try, they are always coming up with new updates

  • @zhenyapestun
    @zhenyapestun 8 днів тому

    Cool video! THX! I was beliver that you have to use local soft - cause its serious and logic to do it that way, but after 40+ hours working in SD on my computer - i decided, that those payed services are not so stupid to use😅

    • @taoprompts
      @taoprompts  8 днів тому +1

      It's worth paying for the convenience sometimes 👍

  • @voyage-digital
    @voyage-digital Місяць тому

    Amazing, thank you!

  • @deimian1111
    @deimian1111 Місяць тому +1

    Excellent Tao. Thank you very much. In a next video, could you show how to create an environment with different of point of view you? Do you think it is possible? As for the character, we could have a living room with furniture, but the point of view will change according where the camera is.

    • @taoprompts
      @taoprompts  29 днів тому

      You can get a background in the same style from different angles, but the exact elements will be different. You'd need real 3D understanding to get the same exact environment from different angles.

  •  Місяць тому +1

    Excellent!
    👏🏻👏🏻👏🏻👏🏻

  • @fabriziorota2671
    @fabriziorota2671 Місяць тому

    Top! Thanks man!

  • @fabiotgarcia2
    @fabiotgarcia2 4 дні тому +1

    Is it possible to put multiple characters on the same image?

    • @taoprompts
      @taoprompts  3 дні тому

      try this method here: x.com/techhalla/status/1859867967049367620

  • @blogger112233445
    @blogger112233445 6 днів тому

    awesome! How would you compare it to Leonardo Ai for character consistency?

    • @taoprompts
      @taoprompts  6 днів тому

      This method works better than Leonardo

  • @braindrift2735
    @braindrift2735 14 днів тому

    Hey thank you very much very nice video do you have experience with large objects e.g. if I want to train certain product let's say a towel with a certain design would it work?

    • @taoprompts
      @taoprompts  11 днів тому

      If the design is very detailed, the AI won't be able to get all the small details correct. But larger patterns and shapes work

  • @Paradoxical66
    @Paradoxical66 Місяць тому

    I can't say how many weeks and even months of videos I have watched and image making and videomaking programs I tried. I have watched some other videos of Open Art recently and came to yours and at the end you mentioned the video making program which is the best I have seen by far and least expensive. I have already done a test using a Pixar image I made using Flux and I find that so far I can get consistent characters using Flux without Open Art. What am I missing? Excellent video by the way and subscribed.

    • @taoprompts
      @taoprompts  29 днів тому

      Thanks for subbing! I'm not sure what you mean. There are many different platforms that allow you to finetune a Flux model. Open art is just one of them that I think works pretty well (and fast).
      But I'd say right now, finetuning a flux model is the best way to get consistent characters, regardless of what platform you use.

  • @emersonfernandes6774
    @emersonfernandes6774 23 дні тому

    Hi Tao, this is all amazing! But I have some questions! I would like to know after the reference sheet is built, can I use any other image AI to create more images of the character I created? Or do I have to create all the images in Openart? My difficulty is also how to transfer the training to Midjourney?

    • @taoprompts
      @taoprompts  23 дні тому +1

      Sure, you can create a character reference sheet in Midjourney and train the custom Flux model in open art

    • @emersonfernandes6774
      @emersonfernandes6774 23 дні тому

      @@taoprompts thanks so much!

  • @TeemuKetola
    @TeemuKetola 26 днів тому

    How about the backside of the character? Without any back view pictures, doesn't the lora and eventual generated images be hallucinations and differ by each generation? Thanks for the great video!

    • @taoprompts
      @taoprompts  24 дні тому

      for human characters it does a decent job with different views, including the back. There will be some variations though. For nonhuman characters it will struggle more to keep the shape

  • @fabiotgarcia2
    @fabiotgarcia2 4 дні тому

    Also, is it possible to train a car and a model then use them in the same image?

  • @voyage-digital
    @voyage-digital 29 днів тому

    It's bought my friend 🧨💥

    • @taoprompts
      @taoprompts  29 днів тому

      I hope it works well for you! There's a ton of different features in OpenArt and they are always updating

    • @voyage-digital
      @voyage-digital 29 днів тому

      ​@@taopromptsIt's crazy stuff. A real ai toolbox. It's perfect, really. I have already made 2 models today 🎉🎉

  • @ArtificialGeekStudio
    @ArtificialGeekStudio Місяць тому

    great work :) do you know how to 2 consistent characters in one scene or a consistent scene? with any ai tool or midjourney

    • @taoprompts
      @taoprompts  Місяць тому +1

      Here's a tutorial that demonstrates 2 consistent characters in Midjourney: ua-cam.com/video/PSges0N_DiE/v-deo.html

    • @ArtificialGeekStudio
      @ArtificialGeekStudio Місяць тому

      @@taoprompts thx

  • @neelimasridharvlogs
    @neelimasridharvlogs 29 днів тому

    Usefull information from India

  • @Steger13
    @Steger13 Місяць тому

    hi, is this platform the best to create same character ? when creating our model do we need full body from every positions too ?thanks

    • @taoprompts
      @taoprompts  Місяць тому

      There are many different platforms that also have this same technique. I'd say they are fairly comparable, this one is easy and fast to use based on my tests. It is helpful to have fullbody shots if you want to generate images with the entire character in them. Or you can just use the face

  • @biy5385
    @biy5385 Місяць тому +1

    honestly how long did it take to make the sample video at the end?

    • @taoprompts
      @taoprompts  Місяць тому

      10-15 minutes. I have a lot of experience with this though

  • @TheStoriesBook-ht4cl
    @TheStoriesBook-ht4cl 27 днів тому

    ❤❤❤❤ awesome

  • @asdweqwerrr-kx8bl
    @asdweqwerrr-kx8bl Місяць тому

    Thanks ❤❤❤❤😊

  • @ToddKauffman
    @ToddKauffman Місяць тому

    But how did you get the initial full body standing rotation poses of the woman in a consistent outfit?

    • @taoprompts
      @taoprompts  Місяць тому

      I showed the prompt in the video

  • @Sean43322
    @Sean43322 Місяць тому

    Always wonderful. 💥❤
    Is it possible to train a model on products like car with this ai?

    • @taoprompts
      @taoprompts  Місяць тому

      Yeah definitely, in the model training page you can choose to fine-tune a Flux model for objects

  • @Bill-ai
    @Bill-ai Місяць тому +1

    Does Flux Dev allow commercial use?

    • @taoprompts
      @taoprompts  Місяць тому

      OpenArt has the commercial license to the Flux models

  • @PokeGamer-jz2fe
    @PokeGamer-jz2fe Місяць тому +1

    Please make a tutorial with Leonardo ai for consistant character ( I have a problem in Leonardo ai and the problem is the character hight and the back pag he wear is change so please make a tutorial.I keep continue to comment until you make the tutorial

    • @taoprompts
      @taoprompts  Місяць тому +2

      Leonardo AI won't work as well as finetuning your own lora/flux model, it's not meant to keep that level of consistency.

  • @bcraigcraig4796
    @bcraigcraig4796 15 днів тому

    what is the correct formal for getting the best prompts using openart ai thanks

    • @taoprompts
      @taoprompts  7 днів тому +1

      that would depend on your image model, for flux similar prompts as midjourney seem to work pretty well

  • @Tales_by_Chommzy
    @Tales_by_Chommzy 29 днів тому

    What about the motion.... You didn't mention the one you used

    • @taoprompts
      @taoprompts  29 днів тому

      here's a tutorial for motion: ua-cam.com/video/3OZKZ20KM08/v-deo.html

  • @iso-
    @iso- Місяць тому

    what plan would i need to create a 3 minute video using these images. i seem to run out in a day with every paid model i use.

    • @taoprompts
      @taoprompts  Місяць тому +1

      depends on how many images you need for the video, I'd say the starter plan should give you enough to create a music video

  • @DrOrion
    @DrOrion Місяць тому

    Can you download the model (is it a lora model)?

    • @taoprompts
      @taoprompts  Місяць тому

      It is a LoRA, I don't think you can download it

  • @cte
    @cte Місяць тому

    The sign up/in buttons don’t work for me. I’m on an iPad. Will try later.

    • @taoprompts
      @taoprompts  Місяць тому

      hmm, I used chrome for this demo. Hope you can get it working somehow

  • @MichaelBaynana
    @MichaelBaynana Місяць тому

    hey tao can i train flux etc with for example animals and anime characters etc?

    • @taoprompts
      @taoprompts  Місяць тому

      Yeah, Flux can work with non human subject and different art styles

  • @noobclown2910
    @noobclown2910 18 днів тому

    Very cool video, but i wish there was a video just dedicated to us who have no cash at all. right from training a model for free to creating videos freely

    • @taoprompts
      @taoprompts  18 днів тому

      If you want high quality results for free, the way to do that is invest in a good GPU and run everything locally

  • @PokeGamer-jz2fe
    @PokeGamer-jz2fe 29 днів тому +1

    We can't train model in free plan 😢

  • @MBop-xh8gl
    @MBop-xh8gl Місяць тому

    Can´t download PDF for free. Site adds a sum and demands card credentials.

  • @Joshua-ju5uc
    @Joshua-ju5uc Місяць тому

    your videos are very helpful. I wonder if you can help me with a problem. I'm trying to use AI to help me make a Fan comic of an old tv show I liked. The show is Teen Titans. The main problem I run into is the cloths and Cyborgs unique machine parts. I'm trying to make the comic feel like it could've been an episode cut up into comic form. Trying to find an art engine to match to style of art for the show is also challenging

    • @taoprompts
      @taoprompts  29 днів тому

      If you use an image from the original show as a style reference, the Ai will be able to replicate the visual style much more closely

  • @HarryClipzFilmz
    @HarryClipzFilmz 9 днів тому

    Can you have consistent scenes too?

    • @taoprompts
      @taoprompts  4 дні тому +1

      I will try and see if it's possible! As far as I know it's hard to get the background details to stay the same for different photos though

    • @HarryClipzFilmz
      @HarryClipzFilmz 4 дні тому

      @@taopromptsI hope they can come up with a way to train environments the same way we can train our character models and also implement video to video so we can make videos like warp fusion

  • @Aryaxis
    @Aryaxis 4 дні тому

    Does this only works, when there is only one char in the pic?

    • @taoprompts
      @taoprompts  3 дні тому

      for more characters try this method: x.com/techhalla/status/1859867967049367620

  • @Rose-hr1nx
    @Rose-hr1nx 8 днів тому

    How you make them moving like that. Like that girl reading a book.

    • @taoprompts
      @taoprompts  7 днів тому +1

      I used the Minimax video generator, here's a prompt tutorial for that ua-cam.com/video/fgZXMKHv78Q/v-deo.html

    • @Rose-hr1nx
      @Rose-hr1nx 5 днів тому

      @taoprompts thank you

  • @CyberPhonkMusic
    @CyberPhonkMusic Місяць тому

    Did you lip sync or just put it on the prompt for him to speak?

    • @taoprompts
      @taoprompts  Місяць тому +1

      Minimax tends to animate speaking even if you don't ask for it, I didn't prompt for any speaking here

    • @CyberPhonkMusic
      @CyberPhonkMusic 29 днів тому

      @@taoprompts thanks bro, amazing tutorials

  • @BrahmandaBytes
    @BrahmandaBytes Місяць тому

    how to do multiple characters in one scene/frame?

    • @taoprompts
      @taoprompts  29 днів тому

      here's a tutorial for doing that in Midjourney: ua-cam.com/video/PSges0N_DiE/v-deo.html

  • @AnneBonny-e6y
    @AnneBonny-e6y 14 днів тому

    is it flux or openart? I'm confused...

    • @taoprompts
      @taoprompts  14 днів тому +1

      flux is the Ai image model, openart is the website/platform where you can access the Ai image model (flux). There are many platforms from where you can access flux

    • @AnneBonny-e6y
      @AnneBonny-e6y 14 днів тому

      @@taoprompts cool! thanks man! I love your channel!

  • @Karnetgyan
    @Karnetgyan Місяць тому

    2nd view from india❤

  • @Larimuss
    @Larimuss Місяць тому

    The best is lora or finetune. But its very time consuimg.

    • @taoprompts
      @taoprompts  Місяць тому

      In my experience LoRA's are pretty fast to finetune, less than 30 minutes or so

  • @levenin9689
    @levenin9689 25 днів тому

    Why does you always show paid sites...we're looking for free sites or with daily credit limits...

  • @SufyanBloch-vy7ps
    @SufyanBloch-vy7ps Місяць тому

    Is there a free method?

    • @taoprompts
      @taoprompts  Місяць тому

      Sure, you can run things locally on your own GPU and finetune a flux model there

    • @SufyanBloch-vy7ps
      @SufyanBloch-vy7ps Місяць тому

      @taoprompts thanks make a video on the topic

  • @bleachedout805
    @bleachedout805 21 день тому

    I needed a tutorial of this with an American accent lol. Too many Germans. I can't understand most of them.

    • @taoprompts
      @taoprompts  19 днів тому

      Glad this helped! Flux finetuning is the best way to go right now

  • @markmacleod2145
    @markmacleod2145 Місяць тому

    Hi Tao - thank you for all your AI youtube videos. Would you be interested in working with me on creating some animated music videos for children's songs as a paid project. Regards - Mark M

    • @taoprompts
      @taoprompts  Місяць тому

      Sure thing, if you want to connect you can message me through my instagram instagram.com/taoprompts/ or my email in my channel info

  • @khawajabashir5834
    @khawajabashir5834 Місяць тому

    you create video about adobe firefly ai video tool

    • @taoprompts
      @taoprompts  Місяць тому

      I will when it's publicly released

  • @mishalshan9271
    @mishalshan9271 Місяць тому

    Only working upgrade your own model

    • @TrunxKraft3000
      @TrunxKraft3000 Місяць тому

      What?

    • @mishalshan9271
      @mishalshan9271 Місяць тому

      ​@@TrunxKraft3000As mentioned in the video, if you want to use it, you have to pay, not free.

  • @audiogus2651
    @audiogus2651 28 днів тому +3

    These are not truly consistent.

  • @themightyflog
    @themightyflog Місяць тому

    Midjourney character ref sucks.

    • @taoprompts
      @taoprompts  29 днів тому

      We'll see if MJ updates their character feature to be more competitive with flux