Slash commands

AIYA is a slash command bot. Here's the available list of commands and information about their respective options.

/draw

The primary command for creating AI art. You'll use this for doing text-to-image and image-to-image commands.

• Prompt - This is where you'll start. Input the things you want to see in the output image.
• Negative prompt - Input things here to try to remove them from the output image.
• Data model - Allows selection from any configured models.
• Steps - The amount of iterations AIYA will go through before completing an image.
• Width - How wide the image will be.
• Height - How tall the image will be.
• Guidance scale - This is how closely AIYA will pay attention to your prompt. Higher = pays more attention. Lower = more freedom to let the model take over.
• Sampler - You can select from one of the sampling methods.
• Seed - This value serves as a basis the generated image. Using the seed can help reproduce the same image or similar images.
• Style - This will list the preset styles that have been configured. Styles are like packages of prompts, shortcuts to save on typing. To see what a style contains, click Styles list after using /tips command.
• Facefix - Tries to fix faces in images using one of the selected available facefix AI models.
• High-res fix - When creating high-resolution images, sometimes undesired results appear (like duplicated anatomy). This tries to fix that by generating the image at a lower resolution (less prone to issues), then upscaling it. The options given are various upscaler models.
• CLIP skip - Some models perform better with different values of CLIP skip.
• Hypernetwork - This can apply a hypernetwork model that can influence the generated image.
• Strength - Used for init_img/init_url. Higher value = AIYA is allowed more freedom to alter the source image.
• Init_img - This will take an image attachment as a starter point to generate an image.
• Init_url - Same as above, but for a URL containing an image.
• Count - This will allow generating multiple images at once using the same parameters.

/identify

This command is used for captioning an image. You can provide an image attachment or URL, and AIYA will generate a caption. There are two options for "phrasing".

• Normal - This uses the CLIP model and will create a standard caption.
• Tags - This uses the deepdanbooru and will give you a description as tags. This could be helpful create prompts on anime-centric models.

/stats

This is a very simple feature. All stats does is count how many images have been generated through /draw. It was the very first extra feature I added! Other stats could be added, but I don't know what would be interesting to have.

/tips

Tips provides a small glossary of information, accessible through its buttons:

• Quick tips - This is a beginner's guide to Web UI's prompting. I wrote this a long, long time ago to help newbies get started with prompting, I've never really updated it.
• Models list - This provides a list of each model configured into AIYA bot. It shows the display name that's normally shown and used, as well as the name of the model (and shorthash) associated with it.
• Styles list - This provides a list of all the styles that have been configured, as well as what words they contain. Currently, there is a bug in which it does not display the negative prompts..
• Hypernets list - This provides a list of each hypernetwork that has been loaded.
• Embeddings list - This lists each of the embeddings that has been loaded. It's categorized into two columns, so people know which can be used for which models (SD 1.x and SD 2.x). That said, AIYA doesn't know which models themselves are SD 1.x or SD 2.x. Bot maintainers will want to have separate Discord channels for each.
• About me - A small credit to myself by linking to this repo. I try not to beat people over the head with credits.

/upscale

This is a simple tool to upscale any existing image. You can feed it an attachment or a URL and resize the image up to 4x. Additional options are provided as follows:

• upscaler 1 - Select among the list of upscaler models loaded. The default is SwinIR_4x if it's installed; otherwise, it's ESRGAN_4x.
• upscaler 2 - This allows to mix in a secondary upscaler model, if desired.
• upscaler 2 strength - This affects how much effect upscaler 2 has (0.0 to 1.0). Setting it to 1.0 will ignore upscaler 1 completely.
• gfpgan- The effect of the GFPGAN face restoration model. It defaults to 0.0, effectively meaning it's unused. Set to 1.0 for the maximum effect.
• codeformer- Same as above, but for the CodeFormer face restoration model. You may use both models simultaneously.
• upscale first - If set to true, AIYA will upscale the image first, then run the face restoration model. The default is false (restore faces, then upscale).