#58438 - safe, generator:invokeai, prompter:siber, tantabus exclusive, twilight sparkle, pony, g4, showcase, solo, tutorial, video, webm

Just a moment

WEBM uploads may take longer to process, it should appear in up to an hour (depending on file size and video length).

Implications might have added tags, so check them in the meanwhile.

InvokeAI is awesome and more people should use it ;) So, I recorded a little video showcasing it. I didn’t want to turn this into a full-on tutorial, but for those unfamiliar with Invoke, here’s a quick rundown of its killer feature - the “infinite canvas”. It’s in the name, really - it is an infinite canvas you can generate on using the bounding box, or as I prefer to call it, the context box. If there’s no raster content in the box - it’s basically txt2img, otherwise it’s img2img. Draw a mask using the masking layer within the context box and you’re inpainting. You can move and scale the box freely around the canvas which allows you to create mood boards, assets, separate characters for further composing, or whatever you please all in one working space.

Another cool thing about Invoke is that it’s like an ogre. It has layers. Four types, to be exact:

Raster - that’s just your plain old pixel data, just like in Krita/GIMP/Photoshop.
Control - special layers for ControlNets.
Regional guidance - think of regional prompter/forge couple or… however you do this in spaghetti UI, but much more convenient. You literally just draw a region and type tags you want to associate with it. Works surprisingly well. Can be both positive/negative.
Mask layer - allows you to draw inpaint masks.

Any of these can be freely enabled/disabled, transformed (move/scale/rotate), copied and converted between types. Like, say you have some old gen of a character with a really good pose/expression or whatever, but you want to place them on a different background or swap out the character entirely, but keep the pose. Just drag the pic into Invoke, cut out the character with the “object select” tool (another cool time saving feature) and convert the layer to ControlNet (appropriate preprocessor will be applied automatically) and you’re good to go.

The only thing that really drags Invoke down is lack of tag autocompletion. One of the devs has shown interest in it, though, so it might be added in the future.

🌟 Check out my CivitAI profile for more AI pony stuff

🌟 Wholesome MLP Mix LoRA <- awesome MLP-focused style for V6

🐙 Check out my GitHub for Pony related AI goodies, namely: 💤 Lazy Pony Prompter and 🎨 PDV6 Styles Collection

Syntax quick reference: **bold** *italic* ||hide text|| `code` __underline__ ~~strike~~ ^{^sup^} _~sub~

Detailed syntax guide

[InvokeAI](https://github.com/invoke-ai/InvokeAI) is awesome and more people should use it ;) So, I recorded a little video showcasing it. I didn't want to turn this into a full-on tutorial, but for those unfamiliar with Invoke, here's a quick rundown of its killer feature - the *"infinite canvas"*. It's in the name, really - it is an infinite canvas you can generate on using the bounding box, or as I prefer to call it, the context box. If there's no raster content in the box - it's basically txt2img, otherwise it's img2img. Draw a mask using the masking layer within the context box and you're inpainting. You can move and scale the box freely around the canvas which allows you to create mood boards, assets, separate characters for further composing, or whatever you please all in one working space.

Another cool thing about Invoke is that it's like an ogre. It has **layers**. Four types, to be exact:
* **Raster** - that's just your plain old pixel data, just like in Krita/GIMP/Photoshop.
* **Control** - special layers for ControlNets.
* **Regional guidance** - think of regional prompter/forge couple or... however you do this in spaghetti UI, but much more convenient. You literally just *draw* a region and type tags you want to associate with it. Works surprisingly well. Can be both positive/negative.
* **Mask layer** - allows you to draw inpaint masks.

Any of these can be freely enabled/disabled, transformed (move/scale/rotate), copied and converted between types. Like, say you have some old gen of a character with a really good pose/expression or whatever, but you want to place them on a different background or swap out the character entirely, but keep the pose. Just drag the pic into Invoke, cut out the character with the "object select" tool (another cool time saving feature) and convert the layer to ControlNet (appropriate preprocessor will be applied automatically) and you're good to go.

The only thing that really drags Invoke down is lack of tag autocompletion. One of the devs has shown interest in it, though, so it might be added in the future.
---
🌟 Check out my [CivitAI profile](https://civitai.com/user/Siberpone/posts) for more AI pony stuff

🌟 [Wholesome MLP Mix LoRA](https://civitai.com/models/509312/wholesome-mlp-mix-pony-v6-xl) <- *awesome MLP-focused style for V6*

🐙 Check out my [GitHub](https://github.com/Siberpone) for Pony related AI goodies, namely: 💤 [Lazy Pony Prompter](https://github.com/Siberpone/lazy-pony-prompter) and 🎨 [PDV6 Styles Collection](https://github.com/Siberpone/pd-styles)

not provided yet

Comments

truekry

Wizzard

I used comfy for like a few weeks when I started playing with AI. Then I switched to Invoke. Been using it for a long time now. Always recommended it if someone asked.

Just a moment

Tantabus

BBCode

Comments