[0:00] in today's video I'm going to show you
[0:02] everything you need to know about the
[0:03] imaged image tool in automatic 1111
[0:06] image to image sketch in painting in
[0:09] paint sketch even in paint upload this
[0:11] video has it all ready let's get started
[0:14] the image to image tab is an essential
[0:16] tool in your air toolbox it lets you
[0:19] create a new image or new elements of an
[0:21] image from an existing picture that you
[0:23] provide there are so many powerful tools
[0:25] here but let's start with the image
[0:27] damage tab itself really you're going to
[0:29] use this tool if you have an image that
[0:31] you want to use as a starting point you
[0:34] can use that image to pull elements of
[0:35] composition and color into a brand new
[0:38] image so I've got an image here that
[0:41] I've generated earlier it's a portrait
[0:43] of a girl on a city street I use the
[0:45] sdxl base model to generate this image
[0:47] but you can use any image you want an
[0:50] existing photo a painting whatever so to
[0:53] start we'll drop that into the image to
[0:54] image box you can also send images here
[0:57] from the text image tab just click on
[1:00] send to image to image under your
[1:02] generation at the top just like we would
[1:04] for text image we can add in positive
[1:07] and negative prompts if you did send
[1:09] your image across from text to image the
[1:11] original prompt will appear just like
[1:13] you see here below the image box you'll
[1:15] see all the settings we can play with
[1:17] now a lot of these settings in this tab
[1:19] are shared with text image so I'll only
[1:21] touch on the image to image specific
[1:23] ones here for a more in-depth
[1:25] explanation of settings check out the
[1:27] video in the description the First new
[1:29] setting is resize mode use this if your
[1:32] new image will have a different size or
[1:34] aspect ratio than your original image
[1:36] for example if your original image is
[1:38] landscape and you want your new image to
[1:40] be portrait the default is just resize
[1:43] this will stretch or shrink your
[1:45] original image to fit the new image you
[1:46] create crop and resize will keep the
[1:49] aspect of your original image and crop
[1:51] it to fit the new canvas size resize and
[1:55] fill takes the original image fits it
[1:57] into the canvas and fills in the blanks
[1:59] with colors from the input image I don't
[2:01] see much use for this one but let me
[2:03] know in the comments if I'm missing
[2:04] something just resize latent upscale is
[2:07] just like just resize but also upscales
[2:10] the image if necessary note that this
[2:12] setting is a bit tricky to get right and
[2:14] there are easier ways to upscale your
[2:16] images below resize modes you'll find
[2:18] settings you're already familiar with
[2:20] including sampling method sampling steps
[2:23] your size and batch settings and the CFG
[2:26] scale the most important setting here is
[2:29] denoising strength stable diffusion
[2:31] makes images by taking random noise and
[2:33] converging it into an image based on
[2:35] your prompt this setting will control
[2:37] how much extra noise is added to the
[2:39] picture and in turn will determine how
[2:41] different the new image is from the
[2:43] original lower settings won't change the
[2:45] image that much and higher settings will
[2:47] make bigger changes I like to use 0.7 as
[2:50] a good starting point okay so now for
[2:52] the fun part once you've sent over an
[2:55] image you can start tweaking some
[2:56] settings to refine your picture we sent
[2:58] this image over from text to image so we
[3:00] have all of our settings and the
[3:02] original prompt now we could just start
[3:04] tweaking our prompt here but bear in
[3:06] mind that the entire prompt plus the
[3:08] image will be used to create a new image
[3:10] so if you aren't using the same model
[3:13] seed and sampler from the original image
[3:15] you may get some weird results so
[3:18] instead we're just going to use some
[3:20] simpler instructions to see how it will
[3:22] affect the image so let's get rid of the
[3:24] prompt and put in pink and and blue hair
[3:30] and let's keep the same sampling method
[3:33] and steps we'll tick restore faces and
[3:37] we'll keep the default on just resize
[3:39] and keep our Dimensions the same and
[3:42] we'll set our denoising strength to 0.7
[3:45] so we should see a bit of a variation in
[3:48] our output foreign
[3:53] seven that's had a pretty big effect so
[3:56] if we want something a little more
[3:57] subtle let's try lowering the denoiser
[4:00] strength to 0.5
[4:03] [Music]
[4:07] and yeah that's a much more subtle
[4:09] impact on the image
[4:11] so you can keep on adding more detail
[4:13] here and reworking things until you get
[4:15] an image that you're looking for
[4:17] when we use this method though we can
[4:19] see that the prompt has had an impact on
[4:22] the entire image the hair and eyes have
[4:24] changed but so has the face and the
[4:26] clothes let's say we just want to change
[4:28] a particular part of the image and keep
[4:31] the rest just as it is that's where in
[4:33] paint comes in in paint is a powerful
[4:35] tool that allows us to paint over the
[4:37] specific parts of images that we want to
[4:39] change this is especially useful when
[4:41] you like the overall composition of your
[4:43] image but you get one of those Cosmic
[4:46] nightmare faces from the hell Dimension
[4:48] just no to start you can click on the in
[4:51] paint tab or just click in paint under
[4:53] the image to copy the image over and all
[4:56] of its settings so now we can just paint
[4:58] over the particular part of the image we
[5:00] want to change in this case the hair we
[5:03] can use the settings in the top right of
[5:05] the box to change the brush size or
[5:07] clear our selection if we go wrong so
[5:10] let's just paint over the hair and then
[5:13] we can check our settings we've seen
[5:14] some of these already so we'll keep
[5:16] resize mode on default mask blur is the
[5:19] amount of blurring that happens around
[5:21] your painted mask lower numbers will
[5:23] give you harder edges while cranking
[5:25] this up will affect how many pixels
[5:27] beyond your mask will be affected by
[5:29] your prompt the default of four is
[5:31] usually fine but feel free to play with
[5:33] this if you're getting hard edges mask
[5:35] mode determines what is actually changed
[5:37] in the image in paint mask will change
[5:39] the parts you've painted while in paint
[5:42] not masked will change everything but
[5:44] the parts you painted this is great if
[5:46] you want to change the background or
[5:48] large amounts of your image in one go
[5:49] next you can choose the masked content
[5:52] setting these settings tell stable
[5:54] diffusion what method it should use as
[5:56] its basis for generating the new image
[5:58] part The Fill option will take the
[6:00] impainted area blur it then generate
[6:03] using that as its base the original
[6:05] setting will use the original image
[6:07] unaltered as its base to generate a new
[6:09] image from latent noise will fill the
[6:12] impainted area with new random noise
[6:14] based on the seed number this can be
[6:16] used if you want to result completely
[6:18] different from the original Late nothing
[6:20] will fill in the painted area with a
[6:22] blend of the colors from the surrounding
[6:24] pixels and base the generation of that
[6:25] these settings will have a dramatic
[6:27] effect on the output but if you're
[6:29] making Minor Adjustments the original
[6:31] setting will usually work great the
[6:33] in-paint area setting is easily confused
[6:35] with mask mode but they are different
[6:37] choosing whole Picture Tells stable
[6:39] diffusion to use the whole picture as
[6:41] inspiration for the in-paint generation
[6:43] while only masked will treat the Mast
[6:46] area in isolation and not draw from the
[6:48] rest of the image typically whole image
[6:50] is the best setting if you want the
[6:52] painted area to blend in but if you do
[6:54] choose only masked you can change the
[6:56] padding size right next to it which will
[6:58] tell stable diffusion how many
[6:59] neighboring pixels it should take into
[7:01] consideration as inspiration for the new
[7:03] generation and of course below that we
[7:05] have the CFG scale and the noising
[7:07] strength let's stick this on 5.5 and 0.6
[7:11] and see what happens
[7:17] now that looks pretty good but we can do
[7:19] even more within painting sketch so
[7:21] let's say we want to give our model a
[7:23] scarf we can send this image to in-paint
[7:25] sketch just like we did before we can
[7:27] paint over the image but we now have the
[7:30] option of choosing colors so let's go
[7:32] with red here and just paint in the area
[7:35] we want to add our scarf to
[7:38] and in the prompt we'll just type in red
[7:40] Woolen scarf
[7:42] for this you usually have to bump up the
[7:45] denoising strength so let's go to 0.8
[7:47] and we can keep all of the other
[7:49] settings the same as before
[7:51] and let's hit generate
[7:56] and yeah that looks pretty good for a
[7:58] first try now you can keep iterating and
[8:01] adding to this to really get the results
[8:02] that you want the last in painting tool
[8:05] is in paint upload now I'm not going to
[8:07] go into too much detail because it's
[8:09] more of an advanced topic but basically
[8:11] you can create a mask in another program
[8:13] like Photoshop using black for the parts
[8:16] that you want to keep and white for the
[8:18] parts that you want to change it's
[8:19] actually a pretty powerful tool and if
[8:21] you have the patience to get really
[8:22] detailed it can be really useful
[8:25] and finally that brings us to the sketch
[8:27] tab if you're struggling to get the
[8:29] image you have in your head onto the
[8:31] screen using image to image or other
[8:33] tools you can always flex your creative
[8:35] muscles and draw it out here
[8:37] simply drop in a white or black mask in
[8:39] the Box provided and use the paintbrush
[8:42] to sketch out your idea
[8:43] use color to highlight the details you
[8:46] want to show up and pair that with your
[8:48] prompt and watch your sketch turn into
[8:49] something incredible
[8:51] there is so much more image to images
[8:53] capable of and in my next video I'll be
[8:55] showing you even more ways to level up
[8:57] your art with the image to image tab if
[8:59] you found this video helpful why not
[9:01] leave a like And subscribe to see more
[9:03] videos like this one thank you for
[9:04] watching to the end I'll catch you later
[9:07] foreign