Flux Kontext - 20 Creative Use Cases To Try - Text Based Image Editing

00:23:15
https://www.youtube.com/watch?v=8L_nFZGp-ho

Résumé

TLDRThe video showcases Flux Context, a new text-based image editing tool by Black Forest Labs, which offers advanced features for character consistency and targeted editing. Users can generate multiple characters in a single image and perform iterative edits. The tutorial guides viewers on how to access and utilize Flux Context on platforms like Freepic and Open Art, highlighting its strengths in colorization, style transfer, and restoration of old photographs. It also discusses the limitations regarding image quality and provides tips for enhancing results. Creative use cases are explored, demonstrating the versatility of Flux Context in various editing scenarios.

A retenir

  • 🖼️ Flux Context allows text-based image editing.
  • 🎨 It excels in character consistency and targeted edits.
  • 🔄 Users can generate multiple characters in one image.
  • 💡 Creative uses include restoring old photos and changing seasons.
  • 📈 Image quality can degrade with multiple edits; use upscale feature.
  • 🆓 Free trial available on Open Art with limited credits.
  • 📜 Multi-image reference feature for consistent character generation.
  • ✍️ Change or remove text in images easily.
  • 👗 Fashion try-on feature for clothing items.
  • 📊 Recommended premium subscription for best results on Freepic.

Chronologie

  • 00:00:00 - 00:05:00

    Black Forest Labs has launched Flux Context, a text-based image editing tool that excels in character consistency and allows for multiple character generation in a single image. The tutorial will demonstrate how to access and use Flux Context on Freepic, with a recommendation for a premium subscription for full features, while also mentioning a free trial option on Open Art.

  • 00:05:00 - 00:10:00

    The tutorial continues on Freepic, where users can upload an image and use text prompts to edit it. An example is given with an image of Nikola Tesla, showcasing the tool's ability to colorize and perform targeted edits without losing character identity. The results are compared with other services, highlighting Flux Context's superior performance in retaining character features and natural colorization.

  • 00:10:00 - 00:15:00

    Despite its strengths, Flux Context has limitations regarding image quality, especially when multiple edits are made in one session. Users are advised to limit edits to maintain quality and utilize the upscale feature on Freepic to enhance image resolution. The tutorial emphasizes the importance of honesty about these limitations while still praising the tool's capabilities.

  • 00:15:00 - 00:23:15

    The video explores various creative use cases for Flux Context, including restoring and colorizing old photographs, changing seasons in images, removing elements, and generating mock-ups. The tutorial encourages users to experiment with the tool's features, such as style transfer and character consistency, while also providing tips for achieving better results and accessing additional resources.

Afficher plus

Carte mentale

Vidéo Q&R

  • What is Flux Context?

    Flux Context is a text-based image editing tool that allows for character consistency and targeted editing.

  • How can I access Flux Context?

    You can access Flux Context through Freepic or Open Art.

  • What are some creative uses for Flux Context?

    Creative uses include restoring old photographs, changing seasons in images, and generating product mock-ups.

  • Does Flux Context maintain character identity?

    Yes, Flux Context excels at retaining character identity during edits.

  • What is the limitation of Flux Context?

    The main limitation is image quality, which can degrade with multiple edits.

  • Can I use Flux Context for free?

    Yes, you can try it for free on Open Art with limited credits.

  • What subscription is recommended for Freepic?

    A premium plus subscription is recommended for the best experience.

  • How can I improve image quality in Flux Context?

    You can use the upscale feature in Freepic to improve image quality.

  • Can I change text in an image using Flux Context?

    Yes, you can change or remove text in an image.

  • What is the multi-image reference feature?

    It allows you to upload multiple images to generate consistent characters in one image.

Voir plus de résumés vidéo

Accédez instantanément à des résumés vidéo gratuits sur YouTube grâce à l'IA !
Sous-titres
en
Défilement automatique:
  • 00:00:00
    Black Forest Labs has just released Flux
  • 00:00:02
    Context, which allows you to do
  • 00:00:04
    textbased image editing, and it's
  • 00:00:06
    actually quite similar to GPT40, but in
  • 00:00:09
    my opinion, I think it does a better job
  • 00:00:11
    at oneshot character consistency. You
  • 00:00:13
    can also generate multiple characters in
  • 00:00:15
    a single image. You can do iterative
  • 00:00:17
    editing and targeted editing. So, if you
  • 00:00:20
    just wanted to change a very specific
  • 00:00:21
    part of an image without affecting the
  • 00:00:23
    rest of it, Flux Context allows you to
  • 00:00:26
    do that. So, in this video, I'll show
  • 00:00:27
    you how to access and use Flux Context,
  • 00:00:30
    as well as a whole bunch of other
  • 00:00:31
    creative use cases. So, let's get
  • 00:00:33
    started. So, the tutorial will take
  • 00:00:36
    place on Freepic, and this is the
  • 00:00:37
    platform that's going to allow you to do
  • 00:00:39
    everything that I'm showing you in the
  • 00:00:40
    tutorial, but you can still try some of
  • 00:00:43
    the prompts in this tutorial for free on
  • 00:00:45
    Open Art. If you are using Freepic, I do
  • 00:00:47
    recommend the premium plus subscription
  • 00:00:49
    since it gives you the most bang for
  • 00:00:51
    your buck. Now, if you are going to try
  • 00:00:53
    this for free, you can try it on
  • 00:00:55
    openart.ai. AI, but like I said, it's
  • 00:00:57
    going to be quite limited, but you can
  • 00:00:59
    sign in over here and then they'll give
  • 00:01:01
    you 40 trial credits, which allows you
  • 00:01:04
    to create eight edits with an image. The
  • 00:01:06
    only reason why you'd need a free pig
  • 00:01:08
    subscription, in my opinion, is if you
  • 00:01:10
    want to follow along with the
  • 00:01:11
    multi-image reference section in this
  • 00:01:14
    tutorial that allows you to upload
  • 00:01:15
    multiple images. You can only do that on
  • 00:01:17
    Freepic and not on Open Art. Anyway, I'm
  • 00:01:20
    still, you know, I want to create the
  • 00:01:22
    opportunity for those who want to go the
  • 00:01:24
    paid route and those who also want to
  • 00:01:25
    try it for free. Anyway, let's get
  • 00:01:27
    started. If you're using free pick,
  • 00:01:29
    check the link in the description. It'll
  • 00:01:31
    bring you over to this page. Go ahead
  • 00:01:33
    and sign in. Make sure you have a
  • 00:01:35
    subscription enabled and then click on
  • 00:01:37
    assistant. You want to click on this
  • 00:01:38
    drop down arrow and select Flux Context
  • 00:01:41
    Max. In my opinion, this gives you the
  • 00:01:43
    best results and has the best prompt
  • 00:01:44
    adherence. Now, it's as simple as
  • 00:01:46
    dragging and dropping an image over
  • 00:01:48
    here. And then we can type a prompt to
  • 00:01:49
    make any edits to an image. If you are
  • 00:01:52
    trying this for free using Open Art,
  • 00:01:54
    check the link in the description as
  • 00:01:55
    well. It'll bring you over to this page.
  • 00:01:57
    Sign in. It'll give you 40 credits. That
  • 00:02:00
    allows you to edit eight images. Then
  • 00:02:02
    over here, you want to click on chat.
  • 00:02:04
    Make sure Flux Context is selected. And
  • 00:02:07
    then you can drag and drop your image
  • 00:02:09
    over here. And this is where you'll type
  • 00:02:10
    in, you know, any prompts to make edits
  • 00:02:12
    to an image. Now, I will be continuing
  • 00:02:14
    with the rest of this tutorial on
  • 00:02:16
    Freepick, but you can still try some of
  • 00:02:18
    these prompts on Open Art. You just
  • 00:02:19
    won't be able to do anything with the
  • 00:02:21
    multi-image reference section of this
  • 00:02:24
    tutorial. So, let's continue on
  • 00:02:26
    Freepick. So, of course, we are making
  • 00:02:28
    edits to an image and we need an image.
  • 00:02:30
    So, I'm going to upload an image of
  • 00:02:32
    Nicola Tesla and I'll be making some
  • 00:02:34
    edits to it just by using text and I'll
  • 00:02:36
    show you just how powerful Flux Context
  • 00:02:38
    is. So once your image is uploaded, this
  • 00:02:41
    is where you can give it some
  • 00:02:42
    instructions. So you can essentially
  • 00:02:44
    change the entire style of an image or
  • 00:02:46
    even do targeted edits. So in this case,
  • 00:02:48
    the first thing I'm going to type over
  • 00:02:50
    here is colorize this image. Right? As
  • 00:02:52
    soon as I type that, all I have to do is
  • 00:02:55
    click on this button because I don't
  • 00:02:57
    need to type anything else regarding
  • 00:02:58
    aspect ratios. I click on this button to
  • 00:03:00
    generate it and then the AI is going to
  • 00:03:02
    colorize this image for me. So you can
  • 00:03:05
    see Flux Context has gone ahead and
  • 00:03:07
    added this natural colorization to the
  • 00:03:09
    image which looks amazing. But where
  • 00:03:12
    Flux Context really shines is with
  • 00:03:14
    character identity and preservation. You
  • 00:03:17
    can see that the colorized result still
  • 00:03:19
    looks exactly the same as my source
  • 00:03:21
    image. And this is a big deal because
  • 00:03:23
    other services like GPT40 and Gemini 2.0
  • 00:03:26
    Flash tends to actually change the
  • 00:03:28
    features on someone's face quite a bit.
  • 00:03:31
    So this one does a really good job. But
  • 00:03:33
    now you can take it one step further
  • 00:03:34
    once your image has been colorized
  • 00:03:36
    because that's essentially, you know,
  • 00:03:38
    putting it in a different style. Let's
  • 00:03:40
    say we can now do targeted editing. So
  • 00:03:43
    maybe I can say change his outfit and he
  • 00:03:45
    is now wearing a black puffer jacket. So
  • 00:03:47
    now it's doing targeted editing. The AI
  • 00:03:49
    knows that it needs to select this part
  • 00:03:51
    of the image. And there we go. It now
  • 00:03:53
    gives him a black puffer jacket. So now
  • 00:03:55
    Nicola Tesla is living within the 21st
  • 00:03:58
    century. And then you can continue from
  • 00:04:00
    here as well, right? You can see the AI
  • 00:04:03
    understands the previous context. So now
  • 00:04:05
    it can take this image and I I can say
  • 00:04:07
    make him wear wear black sunglasses and
  • 00:04:10
    then it's going to do additional
  • 00:04:11
    targeted editing. You'll notice that it
  • 00:04:13
    didn't affect any other parts of the
  • 00:04:14
    images uh of the image but just added in
  • 00:04:17
    those black sunglasses. And this is a
  • 00:04:19
    big deal because this just makes it so
  • 00:04:22
    easy to make changes or any edits to an
  • 00:04:24
    image just by using natural language.
  • 00:04:27
    So, it's only fair to compare these
  • 00:04:29
    results across three services that allow
  • 00:04:30
    you to edit images just by using text.
  • 00:04:33
    Now, I've gone ahead and used this image
  • 00:04:35
    of Nicola Tesla, and I think it's quite
  • 00:04:37
    obvious who the winner is over here.
  • 00:04:39
    It's definitely Flux Context. It's done
  • 00:04:41
    the best job at retaining the
  • 00:04:42
    character's identity, but the
  • 00:04:44
    restoration and the colorization looks
  • 00:04:46
    very natural. With Google Gemini 2.0
  • 00:04:48
    Flesh, I don't like the results of the
  • 00:04:50
    colorization. And then GPT40 tends to
  • 00:04:53
    struggle with retaining character
  • 00:04:55
    identity. And it's also added its
  • 00:04:57
    infamous yellow tint or sepia tint onto
  • 00:05:00
    the image. So flux context is amazing,
  • 00:05:03
    but there is a limitation that people
  • 00:05:04
    are just not talking about, and it's the
  • 00:05:06
    image quality. You'll notice your
  • 00:05:08
    results actually don't look very high
  • 00:05:10
    quality. They don't look that crisp. Uh
  • 00:05:12
    so whenever you guys are editing a
  • 00:05:14
    image, I would recommend that you don't
  • 00:05:16
    make more than six edits within a single
  • 00:05:18
    context window because the quality of
  • 00:05:20
    those images are going to degrade over
  • 00:05:22
    time. Instead, just save an image, then
  • 00:05:25
    create maybe a new chat and make edits
  • 00:05:27
    within a brand new window. But keep in
  • 00:05:29
    mind that image quality degradation is
  • 00:05:32
    definitely going to happen. Now, with
  • 00:05:34
    Freepick, there is a way to get around
  • 00:05:36
    this. Uh they have a feature over here
  • 00:05:38
    which is upscale. So, with this image,
  • 00:05:40
    if I click on upscale, it brings me over
  • 00:05:42
    to this page, and I can upscale this
  • 00:05:43
    image to improve the quality using
  • 00:05:45
    Magnafi. with my creativity. Well, with
  • 00:05:48
    all of these sliders, I leave it on
  • 00:05:49
    default. But I always make sure my
  • 00:05:51
    creativity is minus because if this is
  • 00:05:53
    too high, it'll end up changing the
  • 00:05:55
    image too much. I don't even need to
  • 00:05:57
    type in a prompt. I can simply click on
  • 00:05:58
    upscale and I'll get an image with much
  • 00:06:00
    better quality. So, I really hope that
  • 00:06:03
    Flux improves this uh in the short term
  • 00:06:05
    so that we can get much better image
  • 00:06:07
    quality with our results. But I had to
  • 00:06:10
    mention that because I'm all about
  • 00:06:11
    honesty on this channel. So this is
  • 00:06:14
    going to be the upscale result which of
  • 00:06:16
    course makes the image look a lot
  • 00:06:17
    better. So at least we do have a
  • 00:06:18
    solution for this if you want your
  • 00:06:20
    images to look higher quality. But of
  • 00:06:22
    course Flux needs to ensure that we can
  • 00:06:24
    get highquality results straight out of
  • 00:06:26
    the gate. Anyway, I still think that
  • 00:06:28
    Flux context is absolutely amazing and
  • 00:06:30
    I'm going to add some additional value
  • 00:06:32
    to this video by showing you a whole
  • 00:06:34
    bunch of creative use cases that you can
  • 00:06:39
    try. Hello awesome human. Are you ready
  • 00:06:42
    to see some creative use cases? Let's
  • 00:06:44
    go. Flux Context, in my opinion, does
  • 00:06:47
    some of the best ones restoration and
  • 00:06:50
    colorization for old photographs because
  • 00:06:53
    not only does it retain the person's
  • 00:06:55
    identity, but it also fixes other
  • 00:06:58
    imperfections like scratches or tear
  • 00:07:00
    marks that might be present on that
  • 00:07:01
    image. And then it even goes ahead and
  • 00:07:03
    colorizes that image to give it some new
  • 00:07:06
    life. And the prompt for this is very
  • 00:07:08
    simple. So when you upload a old
  • 00:07:10
    photograph, you just type in restore and
  • 00:07:12
    colorize this image. Remove any
  • 00:07:14
    scratches or imperfections and then
  • 00:07:16
    watch the magic unfold as it restores
  • 00:07:19
    and colorizes these images for you.
  • 00:07:22
    Remember that you're not just limited to
  • 00:07:24
    people. You can also restore and
  • 00:07:25
    colorize locations. Now, you can take
  • 00:07:28
    this one step further and use a service
  • 00:07:30
    like Clling. So I'm using Cling 2.1 and
  • 00:07:32
    I'm doing image to video and I'm just
  • 00:07:34
    typing in a simple prompt like a woman
  • 00:07:36
    smiles and waves. And now it's actually
  • 00:07:38
    going to go ahead and bring that image
  • 00:07:40
    to life. So, not only have we restored
  • 00:07:43
    and colorized an old photograph, but now
  • 00:07:46
    we've also brought it to life. And this,
  • 00:07:49
    in my opinion, is incredible. So, here's
  • 00:07:51
    another result where she's just having a
  • 00:07:53
    really good
  • 00:07:55
    laugh. If you're not doing full
  • 00:07:57
    restoration, you can also just colorize
  • 00:07:59
    a black and white image. So, I had to
  • 00:08:01
    try this with an image of Marilyn Monroe
  • 00:08:03
    as well as Nicola Tesla to see how well
  • 00:08:05
    it actually performed and I think it did
  • 00:08:07
    a fantastic job. The character identity
  • 00:08:10
    has been preserved and the colorization
  • 00:08:12
    looks very natural. You can change the
  • 00:08:15
    time of season in an image. So, I
  • 00:08:16
    started with this image of the Mona Lisa
  • 00:08:18
    in a park during autumn. I uploaded that
  • 00:08:21
    image and then I just simply typed in
  • 00:08:22
    change the weather to a snowy day in
  • 00:08:24
    winter. And the end result looks
  • 00:08:26
    amazing. It's retained the structure as
  • 00:08:28
    well as retaining the character's
  • 00:08:29
    identity. But now it's added in all of
  • 00:08:31
    these piles of snow. It's added
  • 00:08:33
    snowflakes onto her jacket and even onto
  • 00:08:35
    her hair. And now that you have two
  • 00:08:37
    images with a similar structure, you can
  • 00:08:39
    actually generate a very cool video. So
  • 00:08:42
    once you have both of your images, you
  • 00:08:43
    can actually generate an awesome video
  • 00:08:45
    like this that shows you the seasons
  • 00:08:47
    changing. So to do this, I'm using a
  • 00:08:49
    service called Pixver. And on Pixver,
  • 00:08:51
    there's a feature called transision. So
  • 00:08:53
    this allows you to use key frames. So
  • 00:08:56
    you can see over here you can upload a
  • 00:08:57
    first frame and a last frame. So my
  • 00:08:59
    first frame is autumn. My last frame is
  • 00:09:01
    winter. I then type in a simple prompt
  • 00:09:03
    like this. I choose my duration and
  • 00:09:05
    resolution and then click on create. And
  • 00:09:08
    I get this amazing video where I
  • 00:09:10
    actually get to see those seasons
  • 00:09:14
    changing. You can also remove elements
  • 00:09:16
    in an image. So this is also a great way
  • 00:09:19
    to get rid of distractions. So, for
  • 00:09:21
    example, with this image with the Mona
  • 00:09:22
    Lisa, I just simply typed in remove the
  • 00:09:25
    woman in this image, and it removed her.
  • 00:09:27
    It's literally that simple. Here's
  • 00:09:29
    another example with these two birds on
  • 00:09:31
    this branch, and I just typed in remove
  • 00:09:34
    the two birds off the branch, and that's
  • 00:09:35
    exactly what it did. It made sure the
  • 00:09:37
    rest of the image remained untouched,
  • 00:09:39
    but it just removed what I requested.
  • 00:09:42
    So, this is a fantastic feature. This
  • 00:09:44
    means that you can get creative with
  • 00:09:46
    this. So, I'm intentionally removing
  • 00:09:48
    elements from an image in order for them
  • 00:09:50
    to reappear in a very creative way. And
  • 00:09:53
    the way I'm achieving this is with a
  • 00:09:55
    platform called PA because with PA
  • 00:09:57
    there's a feature called PA frames which
  • 00:09:59
    allows me to upload key frames. So, my
  • 00:10:02
    start frame is the image that does not
  • 00:10:04
    contain the birds and my last frame is
  • 00:10:06
    the image that contains the birds. And
  • 00:10:08
    then I simply just type in a prompt
  • 00:10:10
    describing how I want these birds to
  • 00:10:11
    appear on the branch. I choose a
  • 00:10:13
    duration. I click on generate and the AI
  • 00:10:16
    works its magic to create this really
  • 00:10:18
    cool end result where these birds just
  • 00:10:20
    appear out of a puff of smoke. So just
  • 00:10:23
    have fun with this technology. Think
  • 00:10:25
    outside the box and get creative. Flux
  • 00:10:28
    Context handles style transfer
  • 00:10:30
    exceptionally well. You upload your
  • 00:10:32
    image and then you just type a prompt
  • 00:10:34
    describing what style you want to turn
  • 00:10:35
    it into. So maybe you want to see it in
  • 00:10:37
    the style of clmation or in the style of
  • 00:10:39
    a '90s anime. You just simply specify
  • 00:10:42
    that. And because Flux Context does a
  • 00:10:44
    very good job with retaining the overall
  • 00:10:46
    structure and even character identity,
  • 00:10:48
    the end results look amazing. And here's
  • 00:10:51
    something else you can maybe try with
  • 00:10:52
    the style transfer, especially if you've
  • 00:10:54
    generated something that maybe looks
  • 00:10:56
    like a sculpture, like it's made out of
  • 00:10:57
    gold or marble. You can take this over
  • 00:10:59
    to cling using 2.1 and you can type in a
  • 00:11:03
    360° camera orbit around the statue. And
  • 00:11:06
    then, of course, it just makes it look
  • 00:11:07
    more interesting and more dynamic. So
  • 00:11:09
    now you've got this rotation uh with the
  • 00:11:11
    statue. And it's amazing that it can
  • 00:11:13
    turn that image into something that
  • 00:11:15
    certainly looks like it's been sculpted.
  • 00:11:18
    Changing a hairstyle is very simple. You
  • 00:11:20
    just need to specify what color the hair
  • 00:11:22
    needs to be as well as the hairstyle
  • 00:11:24
    that you want to be applied to your
  • 00:11:26
    character and the AI will work its
  • 00:11:28
    magic. I can then go ahead and even
  • 00:11:30
    prompt in a different item of clothing.
  • 00:11:32
    So maybe I want her to wear a beige
  • 00:11:34
    trench coat. So, I'll upload that image
  • 00:11:36
    I just generated where I changed her
  • 00:11:38
    hair color. And then I can type in
  • 00:11:40
    something like make her wear a beige
  • 00:11:42
    trench coat. And then Flux Context does
  • 00:11:44
    a very good job with making sure that it
  • 00:11:46
    doesn't affect other elements within the
  • 00:11:48
    image, but it just changes her item of
  • 00:11:50
    clothing. You can then go ahead and even
  • 00:11:52
    prompt in accessories like sunglasses
  • 00:11:54
    and it does a very good job. You can
  • 00:11:57
    also prompt different facial
  • 00:11:58
    expressions. So, with this example, I
  • 00:12:00
    just simply typed in the woman has an
  • 00:12:02
    expression of excitement. And you can
  • 00:12:04
    see it keeps everything else consistent
  • 00:12:06
    while just affecting her facial
  • 00:12:07
    expression. So, if you want them to be
  • 00:12:09
    sad or maybe they're angry, just simply
  • 00:12:11
    prompt it and it will go ahead and
  • 00:12:13
    change that facial expression for you.
  • 00:12:15
    You can change the text on an image, and
  • 00:12:17
    in my opinion, this is extremely useful
  • 00:12:19
    because not only is it going to match
  • 00:12:21
    the overall typography and font that's
  • 00:12:23
    used on that design, but it's also going
  • 00:12:25
    to remain legible, which is a major
  • 00:12:28
    plus. You can even change text that is
  • 00:12:30
    obstructed. Sometimes you need to
  • 00:12:32
    specify that the text is behind a
  • 00:12:34
    character, but it does a very good job.
  • 00:12:36
    Maybe you just quickly want to change
  • 00:12:37
    the date on a poster or the text that's
  • 00:12:40
    written by the header or maybe you want
  • 00:12:42
    to see a logo with some different
  • 00:12:44
    typography. You can easily change the
  • 00:12:46
    text just by specifying in the prompt
  • 00:12:48
    what needs to be
  • 00:12:50
    changed. You can easily remove all of
  • 00:12:53
    the text from an image. And I tried this
  • 00:12:55
    with this illustration generated with
  • 00:12:57
    GPT40. And as you can see, it removed
  • 00:12:59
    all of the text, but made sure that the
  • 00:13:01
    illustration was still
  • 00:13:03
    intact. You can easily generate
  • 00:13:05
    mock-ups. So, let's say that you maybe
  • 00:13:07
    have an illustration for a t-shirt
  • 00:13:08
    design, and you want to see what that
  • 00:13:10
    design would look like on a t-shirt. You
  • 00:13:12
    can simply prompt it and then have a
  • 00:13:14
    virtual model wearing your design on a
  • 00:13:16
    t-shirt. It does a very good job. You
  • 00:13:19
    can also maybe see what it would look
  • 00:13:20
    like on a mug or maybe on a poster
  • 00:13:23
    that's essentially placed outside an
  • 00:13:25
    establishment. So, if you want to create
  • 00:13:27
    all sorts of mock-ups using a design
  • 00:13:29
    that you provided with, it's going to do
  • 00:13:31
    a very good job. You can always take
  • 00:13:34
    this one step further. If you've got a
  • 00:13:35
    virtual model that's now wearing your
  • 00:13:37
    t-shirt with your design, bring that
  • 00:13:39
    image to life. So, I'm using Clling 2.1
  • 00:13:42
    image to video. I just type in a simple
  • 00:13:44
    prompt like this. And then I get this
  • 00:13:46
    amazing end result. Uh, which has now
  • 00:13:48
    brought that image to life. Uh, so you
  • 00:13:51
    can do the same with a poster as well.
  • 00:13:52
    Maybe you want to add some life to that
  • 00:13:54
    poster and you can bring it to life. So,
  • 00:13:56
    I would recommend doing image to video.
  • 00:13:58
    It just takes it one step further and
  • 00:14:00
    makes the results look a little bit more
  • 00:14:03
    interesting. You can also easily swap
  • 00:14:05
    out a model. So, if I like the overall
  • 00:14:07
    composition and the pose that a current
  • 00:14:09
    model is in, I can simply swap her out
  • 00:14:11
    with something else. So, I can swap her
  • 00:14:13
    with maybe an Asian model or even a
  • 00:14:15
    green alien. So get creative, but you
  • 00:14:18
    can easily do character swaps in an
  • 00:14:19
    image while making sure that everything
  • 00:14:21
    else still stays consistent. If you
  • 00:14:24
    wanted to, you could also use flux
  • 00:14:26
    context as a way to give you some
  • 00:14:27
    inspiration. Let's say you've actually
  • 00:14:29
    sketched something on a piece of paper
  • 00:14:31
    and you want to see what that image
  • 00:14:32
    would look like as a completed
  • 00:14:34
    illustration. This can maybe give you
  • 00:14:36
    some ideas for how you'll actually go
  • 00:14:38
    ahead and apply color to your image. So
  • 00:14:40
    yes, you can even take a sketched image
  • 00:14:42
    and turn it into a completed
  • 00:14:44
    illustration. Flux Context is capable of
  • 00:14:47
    achieving oneshot character consistency.
  • 00:14:49
    And this is quite a big deal because it
  • 00:14:51
    does a very good job with retaining
  • 00:14:53
    character identity and making sure
  • 00:14:55
    everything else is consistent. So I'm
  • 00:14:57
    using free pick and I'm here under the
  • 00:14:59
    AI image generator. So now I can simply
  • 00:15:02
    go to Flux and select Flux Context Max
  • 00:15:05
    to get the best results. And then I
  • 00:15:06
    simply drag and drop an image over here
  • 00:15:08
    of my character that I want to use. So
  • 00:15:10
    I've got this realistic Mona Lisa and
  • 00:15:12
    she's wearing this black leather jacket.
  • 00:15:14
    So, it could either be an image cropped
  • 00:15:16
    in closer to the face or an image that
  • 00:15:18
    shows the face and a little bit of the
  • 00:15:19
    outfit. And it's going to make sure
  • 00:15:21
    everything over here stays consistent.
  • 00:15:23
    So, this means that you can essentially
  • 00:15:25
    give it instructions like in this case,
  • 00:15:27
    I said, "Show me a side profile of this
  • 00:15:28
    image." And guess what? I've got a side
  • 00:15:30
    profile. Maybe I want to see a rear view
  • 00:15:32
    shot of this person. And it's going to
  • 00:15:34
    generate that as well. And you can see
  • 00:15:36
    everything remains consistent, even the
  • 00:15:37
    background in this case. But the best
  • 00:15:39
    part about this is that you can use it
  • 00:15:41
    for a multitude of different scenarios.
  • 00:15:43
    So I could place this person in a
  • 00:15:45
    completely different location. So now
  • 00:15:47
    you can do storytelling with a
  • 00:15:49
    consistent character. So now you can see
  • 00:15:52
    over here I said place her in a seaside
  • 00:15:53
    town in Italy and she's eating ice cream
  • 00:15:55
    and that's exactly what she's doing. The
  • 00:15:57
    only downside with flux context is the
  • 00:15:59
    image quality. You can see it's not
  • 00:16:01
    crisp. But on on free pick there is a
  • 00:16:04
    workaround for this. You just have to
  • 00:16:05
    click on upscale and then you can
  • 00:16:07
    upscale your image using magnafi. I
  • 00:16:09
    always leave this creativity quite low.
  • 00:16:12
    So minus three is fine. You can then
  • 00:16:14
    type in a prompt over here like the
  • 00:16:15
    woman's eating ice cream. You click on
  • 00:16:17
    upscale and you're going to get an image
  • 00:16:20
    with better image quality. So this is
  • 00:16:21
    the upscaled result, right? So it does
  • 00:16:24
    look better than the previous result. Uh
  • 00:16:27
    but that is just a little bit of a
  • 00:16:29
    limitation when it comes to flux context
  • 00:16:31
    is the image quality. But you can still
  • 00:16:34
    achieve character consistency and place
  • 00:16:36
    them in any environment you can think of
  • 00:16:39
    to tell your own stories. I am going to
  • 00:16:41
    be honest. However, if you are focused
  • 00:16:44
    on, you know, character consistency, I
  • 00:16:47
    do think that runway is better in this
  • 00:16:50
    case. So, this is runway references. It
  • 00:16:52
    functions exactly the same. You'll
  • 00:16:53
    upload a reference image. You can then
  • 00:16:56
    type a prompt, choose resolution, and
  • 00:16:58
    you can generate images. The thing over
  • 00:17:00
    here is that the quality of runway
  • 00:17:02
    reference images just look better. Like
  • 00:17:04
    straight out of the gate, they look much
  • 00:17:06
    better. Uh so if you are focused on
  • 00:17:09
    consistent consistent uh you know
  • 00:17:12
    storytelling with characters I think
  • 00:17:13
    this would be a good alternative until
  • 00:17:16
    you know flux context can actually
  • 00:17:17
    generate images with much better image
  • 00:17:19
    quality straight out of the gauge. You
  • 00:17:20
    can see this does the same thing right I
  • 00:17:22
    can see side profiles and I also have a
  • 00:17:24
    consistent character. So maybe consider
  • 00:17:26
    runway references as well as an
  • 00:17:28
    alternative. But anyway this is one of
  • 00:17:30
    the great features of flux context. You
  • 00:17:33
    can achieve character consistency.
  • 00:17:36
    You can change the time of day. And I've
  • 00:17:38
    noticed that this works particularly
  • 00:17:40
    well with any images where you have a
  • 00:17:42
    landscape and then some architecture
  • 00:17:44
    because you can prompt it to actually
  • 00:17:46
    turn the lights on inside a building. So
  • 00:17:48
    with these examples, I've basically gone
  • 00:17:50
    from daytime to the evening. And you can
  • 00:17:52
    see how it does such a fantastic job
  • 00:17:55
    with turning on those lights and still
  • 00:17:57
    ensuring that everything looks natural
  • 00:17:59
    within the image. So definitely give
  • 00:18:00
    this a try.
  • 00:18:02
    You can virtually add furniture and
  • 00:18:04
    decorations into an empty room using
  • 00:18:07
    Flux Context. And this use case is
  • 00:18:09
    actually very useful. So maybe you've
  • 00:18:11
    even got an image of your own apartment
  • 00:18:13
    and you maybe want to see how it could
  • 00:18:14
    be decorated with different items of
  • 00:18:16
    furniture. Maybe you want some modern
  • 00:18:18
    furniture with a specific color palette
  • 00:18:20
    as well. You can specify those colors in
  • 00:18:22
    the prompt and then just tell it to fill
  • 00:18:24
    the room with furniture. You can
  • 00:18:27
    generate a product advertisement just
  • 00:18:29
    from using your product. And because
  • 00:18:30
    Flux Context does such a good job with
  • 00:18:32
    maintaining the overall consistency and
  • 00:18:34
    structure of whatever item you throw at
  • 00:18:36
    it, it's going to generate a really cool
  • 00:18:38
    poster. And Flux Context also does an
  • 00:18:41
    exceptionally good job with generating
  • 00:18:43
    the text. I've noticed that there are
  • 00:18:45
    hardly any spelling mistakes whenever
  • 00:18:47
    you generate any text within a poster.
  • 00:18:49
    So go ahead, supply it with an image of
  • 00:18:51
    your product and generate some epic
  • 00:18:53
    looking posters. Of course, you can
  • 00:18:56
    always take it one step further. Now, I
  • 00:18:57
    uploaded this poster to Cling and I'm
  • 00:19:00
    using Cling 2.1. And because there's
  • 00:19:02
    these elements of electricity within
  • 00:19:04
    this image, it actually looks quite
  • 00:19:05
    cool. This was the prompt I used. And
  • 00:19:07
    now I brought that poster to life that's
  • 00:19:10
    promoting this energy drink called
  • 00:19:14
    Context. Now, I want to show you
  • 00:19:16
    multi-image references. And remember,
  • 00:19:18
    like I mentioned at the start, you need
  • 00:19:19
    a subscription on Freepick to actually
  • 00:19:22
    access this. So if you're on freepic, go
  • 00:19:24
    ahead and click on these three lines.
  • 00:19:25
    Click on the image generator. It'll
  • 00:19:28
    bring you over to this page. Click on
  • 00:19:29
    mode here by flux. Select flux context
  • 00:19:33
    max. And you'll see that we have a image
  • 00:19:35
    references section. So this will allow
  • 00:19:37
    us to upload two images and we can
  • 00:19:39
    generate those two items in a single
  • 00:19:41
    image. With multi-image reference, you
  • 00:19:43
    can generate images of someone holding a
  • 00:19:46
    product. So maybe they're promoting an
  • 00:19:47
    energy drink or perfume, whatever you
  • 00:19:49
    want them to hold. it it makes this
  • 00:19:51
    process extremely easy. So you need an
  • 00:19:53
    image of a character and then in this
  • 00:19:55
    case I've got a product image of an
  • 00:19:57
    energy drink called context. Now you can
  • 00:20:00
    see over here it's it says it's
  • 00:20:01
    experimental because sometimes you just
  • 00:20:03
    don't get ideal results. If you are
  • 00:20:05
    going to generate images of someone
  • 00:20:06
    holding a product in your prompt you
  • 00:20:08
    always need to specify the actual size
  • 00:20:11
    of the product so that it's not a
  • 00:20:14
    massive product in their hands because
  • 00:20:15
    by default the products tend to be quite
  • 00:20:17
    large. So just always put the size. So
  • 00:20:19
    you can see over here I said small size
  • 00:20:21
    and then you can choose an aspect ratio
  • 00:20:23
    whether you want it to be you know
  • 00:20:24
    widescreen or if you want something to
  • 00:20:26
    be more vertical and then once you have
  • 00:20:28
    that in mind just go ahead and click on
  • 00:20:30
    generate and you can make people
  • 00:20:32
    actually hold products. Here are some
  • 00:20:34
    other examples where I made Nicolo Tesla
  • 00:20:36
    hold a small chibi toy of himself and
  • 00:20:39
    another one where he's even holding a
  • 00:20:40
    toy Tesla. So this is probably useful
  • 00:20:42
    for e-commerce if you want people to
  • 00:20:44
    actually hold real world real world
  • 00:20:47
    products because it does quite a good
  • 00:20:48
    job with retaining the consistency.
  • 00:20:50
    Although sometimes if text is very small
  • 00:20:52
    on a product it can become a little bit
  • 00:20:55
    inconsistent but definitely give this a
  • 00:20:57
    try. You can also do fashion tryon with
  • 00:20:59
    multi-image reference. So, first you
  • 00:21:01
    upload an image of your model. Then you
  • 00:21:03
    upload an image of the garment. And then
  • 00:21:05
    whenever you crafting your prompts, you
  • 00:21:06
    need to specify what type of item of
  • 00:21:09
    clothing is actually present within that
  • 00:21:11
    image. So if it's a jacket, you want to
  • 00:21:12
    specify that it's a jacket. If it's
  • 00:21:14
    baggy pants, you want to specify in the
  • 00:21:16
    prompt that it's baggy pants. I notice
  • 00:21:18
    that this just helps to give you much
  • 00:21:20
    better results. But the overall
  • 00:21:22
    consistency even with that item of
  • 00:21:24
    clothing is very, very good. So this is
  • 00:21:26
    something you can definitely give a try.
  • 00:21:28
    And you know to get past the whole one
  • 00:21:30
    item limitation you can essentially
  • 00:21:33
    generate in stages. So you can maybe
  • 00:21:36
    generate an image where the character is
  • 00:21:37
    wearing pants and then that new image
  • 00:21:39
    becomes your starting image and then you
  • 00:21:41
    can you know upload an image of a jacket
  • 00:21:44
    and then craft a prompt for that to
  • 00:21:46
    create an entire outfit. Uh but you can
  • 00:21:49
    definitely try fashion tryon by using
  • 00:21:51
    reference images. So multi-image
  • 00:21:53
    reference is very useful. You can use
  • 00:21:56
    multi-image reference to generate two
  • 00:21:58
    consistent characters in one image. And
  • 00:22:00
    you can give them all sorts of
  • 00:22:01
    instructions. Like maybe one of them is
  • 00:22:03
    holding a thumbs up, the other one is
  • 00:22:05
    holding up a peace sign. Uh you can
  • 00:22:07
    describe the location, the time of day.
  • 00:22:10
    So have fun with this. And you can put
  • 00:22:11
    people in all sorts of different
  • 00:22:13
    scenarios. By the way, if you guys want
  • 00:22:15
    access to all of these prompts, just
  • 00:22:17
    check the link in the description. I've
  • 00:22:19
    created a Canva file. You can just click
  • 00:22:21
    on the link and you can gain access to
  • 00:22:22
    all of these prompts completely for free
  • 00:22:24
    as well. So enjoy. So that's going to be
  • 00:22:27
    the end of this tutorial. Now you can
  • 00:22:29
    see that Flux Context is capable of
  • 00:22:30
    doing all sorts of amazing things when
  • 00:22:33
    it comes to textbased editing and it
  • 00:22:35
    just makes our life so much easier. So
  • 00:22:38
    feel free to let me know what you think
  • 00:22:39
    down in the comments below. Anyway, you
  • 00:22:41
    guys are super awesome. So stay tuned
  • 00:22:43
    for some more videos and tutorials and
  • 00:22:45
    goodbye.
  • 00:22:48
    Hello awesome human. Thanks for watching
  • 00:22:50
    the video. Don't forget to subscribe for
  • 00:22:52
    more awesome content.
  • 00:22:54
    [Music]
  • 00:23:05
    [Music]
Tags
  • Flux Context
  • image editing
  • text-based editing
  • character consistency
  • colorization
  • targeted editing
  • Freepic
  • Open Art
  • creative use cases
  • image quality