Mastering Inpainting with Invoke to Add, Remove, and Transform Objects In Your Images

[@MujaidinilTalebbanoNapoletano]

I followed it from start to finish and it was interesting and constructive, thanks for the work you do!

[@VirtusRex48]

Can't wait to see the training process!!!

[@autonomousreviews2521]

Fantastic share - thank you :)

[@ExplorewithZac]

I'm currently in the process of building an inference UI very similar to InvokeAI... But the truth is that InvokeAI is so good that sometimes I wonder why I'm even building this... lol The only way mine differs is having a lot of these features automated and it has a Compose, Transform, and Enhance tab. I also have my Models & LoRas dropdown on the right panel above the Image Gallery, because I have found that the right panel tends to have very little content in it... Having models on the right, with images, feels intuitive because they are both assets that are managed by the user.

[@ExplorewithZac]

Compose Workspace = txt2img and img2img Transform Workspace = inpainting, outpainting, live drawing Enhance Workspace = Upscaling. Exposure, contrast and white balance correction.

[@Pawel_Mrozek]

Damn. I was looking where is denoising strength in the new version for larger part of en hour ;) Why it is there no one knows.

[@AndreyJulpa]

So much usefull information, thank you! I have a question, I have a 3d render of interior room. Can I somehow change lightning to night/morning etc. without changing geometry and texture of furniture and other details?

[@MagicBurnAI]

Please fix the issue that with each update i have to download everytime the same model.

[@Kentel_AI]

Thanks :)

[@gamalfarag]

where can i download customillusionxl model ??

[@NB-ec9wc]

where i can download custom illusionXL model, please

[@entoincognito6597]

Hi, thank you for this very interesting tutorial. I'm new so I have a silly question, but is it possible to simply remove the background from an image and export the image as a PNG?

[@invokeai]

You can't export as a transparent png, yet.

[@HyewonAn-q7r]

와 미쳤다..

[@simjounhax]

저도요... ㄷㄷ

[@garry3989]

Can I ask if this is version dependent? The reason I'm asking is I'm running 5.4.1rc1 and there is no denoising slider on the generation tab

[@garry3989]

Ignore that I found it :)

[@joechip4822]

What I don't understand... When you created the images for in-fill and outpaint right at the start - why wasn't the prompt of the night scene applied any more? The prompt was still there!

[@invokeai]

Similar to how the Lower Denoising Strength example from the "Nighttime" example didn't change the image to a night time scene (that only happened at a strength of 1), when infilling/outpainting, it is using the color from the existing image for that context.

[@TheNjordy]

Wait, so 1.0 denoise still leave some "meanings" on the canvas, the prev image is not completely gone?

[@Avalon19511]

Can you guys get Playground 2.5 to work in Invoke?

[@invokeai]

We likely won't be incorporating it, but contibutors are welcome to add it.

[@Avalon19511]

@@invokeai why would they add it if it doesn't work in Invoke which is too bad because personally I think it's one of the best models out there

[@Taladar2003]

I think the sadness might have gone better if you had included the ears in the inpaint mask, that expression in animals is often portrayed with droopy ears.

[@invokeai]

Great point!

[@DezorianGuy]

Could you please use examples with human characters. You only show landscapes and that's maybe not what people mostly generate. Or is Invoke only good at architecture and landscapes?

[@invokeai]

We typically ask our live audience for direction and this is what they typically ask for, so feel free to join a future live stream and make your request!

[@DezorianGuy]

@@invokeai I don't know your audience, but I am quite sure you should include a humanoid example in each of your future videos. Or are there Invoke tutorial channels focusing on real life generating scenarios?

[@TheNjordy]

I'm not sure I like the idea of moving the denoising slider to the layers area. It's still a parameter for the generation process. Perhaps it would be better to place this slider in one of the corners of the canvas box, and if you move the slider, it would be nice if the image inside the canvas box would show increased noise. Why? First, this would make it very easy to understand for a newbie who might not grasp what it does without a lengthy tutorial explanation. Secondly, it's an additional visual aid for the user to see how much information is being approximately destroyed. We are visual, not digital creatures, and it's not easy to imagine adding noise to an image by 11%.