InvokeAI - Unified Canvas Basics

InvokeAI - github.com/inv...
With the Unified Canvas, InvokeAI allows you to leverage Stable Diffusion and derivative models to generate new images, extend and update existing images, and transform your work in entirely new ways.
This video covers these core use cases:
- Getting Started
- Context
- Inpainting
- Image Coloring & New Context Generation
- Outpainting
- The Canvas Toolbar
This software is free and open source - You can download and run it here, but make sure to check system requirements.
github.com/inv...
Join the Discord for help and tips!
/ discord
- Thanks from the InvokeAI Team
----------------
Music by FASSounds from Pixabay
pixabay.com/us...

[@ianjohnson3546]

@0:04 Bounding Box - @0:56 Getting Started - @1:15 Generating an Image - @2:03 Context @2:35 Inpainting - @3:42 Base Layer - @4:19 Mask Layer - @5:23 Bounding Box / Context - @7:20 Image Coloring / fixing the bad results - @9:47 Prompt tweaking / fixing @11:13 Outpainting - @13:05 Fixing the bad results - @14:18 Seam Correction - @18:40 Adding an object @21:15 Toolbar

[@Metia_9898]

Bless you, brother.

[@mastakilla2253]

Thank you so much! It is interesting to find out how to add other models(.ckpt) to Invoke AI. Good luck

[@Mocorn]

Nothing beats invoke AI for creativity. I've been having so much fun with this the last week.

[@vseslavk.8174]

Useful guide, you need more of these

[@ai_intervention]

Thanks for this. Awesome tutorial and even better software! I'm looking forward to seeing Invoke develop more, the UI is already the best IMO.

[@TheSinbio]

If possible for the next release: - Add new models. - Train the model Great job! but it's still green. :D

[@paskmoe4969]

You should do a tutorial on setting up custom trained models...and how to include them in InvokeAI. Frankly I did not make other things work than directly downloaded files from inside InvokeAI.

[@invokeai]

It'll be easier in our next release - You'll be able to do it from the UI! :)

[@antonioedmilson1810]

@@invokeai Thanks for this, i have a problem with "CUDA out of memory", but i already have the requeriments

[@DemShion]

@@antonioedmilson1810 before generating stuff, change your sampler to anything and then back to the sampler you actually want to use, this solves the problem for me

[@ZeroCool22]

Wouldn't be better to work from the start with the "sd-v1-5-inpainting.ckpt" instead regular one?

[@invokeai]

Personal preference, but I don't really like the inpainting model - Mucks with my use of standard canvas tools a bit (Inpainting model specifically ignores a lot of settings)

[@terbospeed]

Bioluminescent Mushroom Umbrella - New band name!

[@epops5]

Separate field for the negative prompt would be great

[@RenewedRS]

Oh my god I had no idea I could turn off snap to grid lol god dammit

[@TakedaTakeishi]

Thanks for the tutorial. Now I can generate powerful images! One question, did you try changing the model from stable-diffusion-1.5 to inpainting-1.5? I tried it for inpaintings and outpaintings, and the results were... wow!

[@trevor3403]

I'm wondering this as well. DO you notice a difference from SD 1.5 vs inpainting 1.5? Which would you recommend?

[@TakedaTakeishi]

@@trevor3403 If you want a new image or very distinct of the you have, use a stable diffusion 1.5; but if you want a image that inspires in the actual image use inpainting 1.5

[@qvisionstudios]

Thanks for the walk through. It was really helpful. Any way we can integrate SD 2.1 as a model?

[@andrewowens5653]

I have stable diffusion running with automatic 1111. But my operating system is Windows 7, because there is some Legacy software I must have access to. Unfortunately I haven't been able to install Invoke-AI. Does anybody know if this is compatible with Windows 7?

[@PetrasL_1976]

Thanks! It was really hepfull. Could you some day describe a bit, what difference between samplers?

[@babyfox205]

Best since sliced bread 🍞 !

[@SaiMorphX]

Sucks I couldn't get just fire around objects, it keeps changing the objects, I mask around it, and all I ask it to do "fire, sparks (fire:2),(sparks:.02)" But I did learn a LOT here thank you!!!! I think the community edition isn't as good as whatever you are using, I don't see any of those options!

[@TVdosPatriotas]

Thank you. This was very helpful.

[@MartinSjolte]

Helpful tutorial, thx but please avoid the background music. The bossa Nova was very distracting and load, I was on a sunny beach in Rio most of the time😏

[@civic.sentinel]

What are the maximum dimensions of the final image you can make with Invoke AI?

[@bloxyman22]

Is it possible to use hypernetworks with this? Got several models I would like to use with InvokeAI

[@Igrom_Gamedev]

when will SD 2.1 support be available?

[@andriiB_UA]

Why? 1.5 more better now

[@xtomvideo]

Very cool. Is it possible to mask something like the bottle and tell it to scale it up or down?

[@darrenelkins5923]

It seems to struggle with upscale images

[@KevinTurner-aka-keturn]

Thanks for showing us another workflow video! Does your recording software have an option to include the cursor in the capture? We were missing it on this one. "Horizontal flat unfocused landscape" is kinda easy-mode for outpainting. Looking forward to other examples in the future!

[@rodrigop.9071]

Thank you very much for creating this! Is there a guide that explains each of the settings?

[@andersonOak]

I'm really enjoying InvokeAI!👏 Thank you very much for the valuable tips.👍 Keep up the amazing work you are all doing.🦾

[@ochjich]

hello, is it possible to use the newest version of sd 2.1? wouldn't it improve the quality of generated content?

[@jordanelbers]

could you do a tutorial for Image to Image? Maybe something like taking a face from one image and creating a new image around it. I haven't been able to get good results with it yet, but I am new to all of this. Thanks! Liked and subscribed!

[@borisrusev9474]

How exactly does the context window affect the masked region? In this specific case, why did you focus only on the bottle? Why didn't you include say the ocean or the sky?

[@kallamamran]

Very good video! Thanks 👍

[@VictorKaido]

will anyone make a step by step video how to use basic tools like enhancing and extending existing images without prompt and with automated process?

[@arifmohamedmeghji]

Good tutorial

[@void2258]

I've tried drawing in things but I never seem to be able to massage the process to get quite what I want.

[@fSharkCom]

Fantastic tool!

[@Seehart]

Took me awhile to figure out that I need to select an inpainting model to do inpainting.

[@deep6293]

In my opinion the best ui for stable difussion. But: img2img isn't working for me. No matter what strength I use, I always get random pictures far away from my initial picture :-(

[@huzaifa.arshad]

In Unified canvas, when I move the bounding box, and generate the image using the same prompt and settings, it will generate the new image and wont generate the extension of the same image. what i am doing wrong?

[@sandiMexicola]

thanks for the video! :)

[@devilhunter8752]

How i can do image training?

[@mrzainozz1686]

Can we get a installation video !

[@asdasdaa7063]

doesnt work with other models for some reason? I am forced to use the default model or it just gets stuck on processing. Wont let me cancel it to, have to restart.

[@szaibot]

HI, i love it :D is somewhere in ui prompt history for unified canvas?

[@ak-gi3eu]

i hope there would be a tool for photobashing layers or sketching tools and prompt it by categories and layers

[@rickscreations3159]

when i am trying to use outpaint it keeps making blurry images and i can not replicate the same steps you are doing in the video i have tried to ask in the discord group but nobody is responding

[@SwissPitBull]

Why you hide the mouse pointer... I have no clue what you are clicking.

[@caglisto]

can someone help me to install? I'm getting this InvokeAI '""' is not recognized as an internal or external command, operable program or batch file.

[@Sh0GaN785]

Has anyone else had issues on the Unified canvas with models generating random patterns? I’ve been trying OpenJourney lately. Had great success with single text-to-image generation but with the canvas (or sometimes in image-to-image) it doesn’t seem to carry over some settings thus producing bad results.

[@andyrevo8081]

Guys, I just installed the recent version 2.2.5 - and only get a blank screen in the browser. Several people have reported that issue. Installion was fine, but launching it not so much. Although it claimed to have started the InvokeAI web server and I entered the correct localhost address, only a white page appeared in the browser . Since I am not the only one - what can we do, or is there a way to use 2.2.4?

[@loldfg]

Does this work with AMD GPU's?

[@CYBERNORM]

20:08, when the AI kicks in!

[@oblvnus]

Can invoke use the new dreamlike weight

[@LuckysLukes]

Can I run this om ny MSI laptop with NVIDIA® GeForce RTX™ 3060 Laptop GPU 6GB GDDR6?

[@havemoney]

Пожалуйста, под AMD, для Windows =)

[@canalluzdivina2424]

Could you work out a single windows executable file with all the require installations scripted to be run? Thank you.

[@belucheez]

Any good model suggestions?

[@ShinAkuma200X]

Can't even get back to this. Invoke is doin that thing where the directory is under Users... but it's not..

[@henrihenro6352]

I'd like to use InvokeAI but instalation always fails...

[@NovaSaintz]

This is a better photoshop

[@SoullessAIMusic]

inpainting seems so cool but it just takes up a tad bit too much vram to work. At least on my rig inpainting takes more vram thn generating a new image.

[@eddiej.l.christian6754]

"everything bagels and cream cheese" what does this do????

[@belucheez]

Installing for this is tough... invoke batch file never shows up. Another time python is apparently not installed correctly... next the weight files can't be found. Anyway this tool rocks from the perspective of someone who can only watch videos of it lol

[@Mocorn]

For me it was one click install, not a single issue anywhere and it works perfectly. I did have automatic1111 installed prior to this so I might have gotten some of the Python dependencies from that already.

[@belucheez]

@@Mocorn Yeah after I posted I stopped being a baby and re did it again. It works good now!

[@Mocorn]

@@belucheez good work, this UI is worth having up and running. It has quickly become my go-to for creative SD work.

[@yuwong5850]

All brushes doesn't work at all

[@gregorycoleman2149]

Does invokai work with amd cards?

[@KevinTurner-aka-keturn]

If you have Linux with ROCm, it should. I haven't heard of any ROCm equivalent for Windows though.

[@aDAMoscar]

number 1 comment me!

[@MODEST500]

Wth 👁️👄👁️ open source and free 😅 ?

[@aryanmittal4663]

Please provide detailed video for installation.

[@jasonstetsonofficial]

Yes stable diffusion does not stand a chance in outpainting...... so shit.. DALL·E 2 for the win in outpainting

[@wycoconutgaming]

Thanks for the video. Am using 3.5.1 and unfortunately we can no longer collapse the image list on the right hand side of the unified canvas for a bigger canvas view. Could you bring it back please??? I have a pretty tiny screen... 🥲

[@wycoconutgaming]

Ah... I discovered I can pull the divider all the way to the right to collapse the image list panel for more viewing space... 😋