Slack Bot Usage
Learn how to use the AppMana Slack Bot to effectively generate images.
Image Generation Guide
Welcome to the guide for using our tool. Here, we will outline the functionality of the bot, the basics of how to prompt for images, and some example prompts for reference. Prompts work best when they are kept simple, allowing the bot to fill in the details creatively. However, if you have a specific composition in mind, you can include explicit descriptions to guide the output.
Generating Images
To generate images, begin by invoking the bot via @AppMana Generate # [landscape/square/resolution] images of... . Each prompt should first outline the scene, focusing on the characters involved, their interactions, and the background or setting. Then, additional details such as the desired lighting/atmospheric effects, as well as the resolution and model to use, can optionally be indicated.
As a generate template, the query could be:
@AppMana Generate [# images] [resolution] images of [Cast and interactions], [background/setting]. [Additional descriptions]. Use the [one shot/sticker] model.
Tips and Tricks
- The cast members should be referred to by their default names, such as "Mr. Monopoly Man", or "Scottie".
- The cast members will be wearing their "classic" outfits if not prompted (for example, Mr. Monopoly Man will be wearing his tuxedo and bowtie). This can be changed by explicitly describing the outfit. For example, "Mr. Monopoly Man, wearing swim trunks"
- If citizens are desired in the scene, prompt by explicitly using the term "Citizens". For example, "In the background, a few citizens are visible wearing warm winter attire"
- The prompts do not have to include any cast members. The tool can be prompted with just a description of a background, give it a try!
- If you'd like to modify the prompt, reply to the chatbot in the thread created for the generated images. For example, you can reply to the chatbot with "Generate these again, but instead use soft, night-time lighting"
Other functionality
Our tool also currently supports upscaling and background removal. These can be directly prompted by replying to the bot in the thread (click "replies" under the bot). For example, if the image labeled as A3 should be upscaled, in the thread, reply with "Upscale the image in A3"
To view the generated photos in a gallery, click the "gallery" button under the generated images in the slackbot.
Example prompts
Generate 8 4:3 landscape pictures of Mr. Monopoly Man and Scottie, walking through the forest next to each other. The lighting should be soft and sunset. Use the one-shot model.Generate 4 square images of Mr. Monopoly Man, wearing a scarf, standing next to three Citizens, all wearing warm winter coats, building a snowman together, in a snowy park in the winter time. Snow covered trees can be seen in the background. The lighting is soft and in the evening. Use the one shot model.
Training Reference Terms
The slack bot will automatically attempt to augment your prompt so that it uses the exact terminology from the training set, but if you want to specifically target a known pose, camera angle, or facial expression, this is the full list.
-
Poses:
- angry
- at the peak of a jump
- both arms outstretched by his side
- cheering
- chin resting on one hand
- finger gun
- fists raised triumphantly
- gesturing a thumbs up
- gesturing with hands
- hand on her hip
- hanging onto the side of an old coal train
- jump
- jumping
- jumping from being surprised
- knees bent
- looking deep in thought
- looking off into the distance
- looking to the side
- mid jump
- motioning to the side
- pointing to the side
- pumping his fist in the air
- running
- running forwards
- showcasing
- standing
- swinging his arms
- talking
- thinking hand on chin
- thumbs up
- walking
- walking forwards
- waving with one arm
-
Camera Angles:
- 18mm shot, 24mm shot, 25mm shot, 26mm shot, 28mm shot, 30mm shot, 32mm shot, 34mm shot, 35mm shot, 36mm shot, 38mm shot, 40mm shot, 42mm shot, 44mm shot, 46mm shot, 48mm shot, 50mm shot, 52mm shot, 54mm shot, 56mm shot, 58mm shot, 60mm shot, 62mm shot, 64mm shot, 66mm shot, 68mm shot, 70mm shot, 80mm shot
- closeup shot
- extreme wide shot
- from the left
- from the right
- full body shot
- full frontal shot
- full shot
- in the left third of the image
- in the right third of the image
- medium shot
- wide shot
-
Facial Expressions:
- angry
- big smile
- eyebrows raised
- eyes closed
- eyes open
- looking downwards
- mouth closed
- mouth open
- no expression
- shocked
- smile
- smiling
- surprised
- surprised look
- tongue out
- winking
Inpainting
Our tool also supports inpainting capabilities for re-generating portions of the image to fix flaws, or replace elements with something different. This functionality has been directly added to the slackbot, and can be accessed via the Inpaint button located under the grid of images. This tool can be used for a variety of use cases, from fixing improperly generated hands and facial features, to replacing background elements. We will provide some simple guidelines below, as well as some examples of what the tool can do, but these are just a starting point!
Tips and Tricks
- When the inpainting workspace is opened, you will see a gallery of images. Click one to begin inpainting.
- With the "eraser" tool selected (default), begin drawing a mask over the portions of the image you'd like to re-generate.
- The caption can be left as is for most cases of fixing erroneously generated elements. This will simply re-generate the masked portion of the image, in the context of the original prompt.
- For more detailed inpainting, the prompt can be modified to explicitly reference the element being added or replaced. In our experience, the style of the image ("3D cartoon rendering style") should always by included in the prompt.
- Clicking on "See Image Gallery" will show all previous inpainting generations, as well as the original images produced. Clicking the left and right arrows will also allow you to cycle through the inpainting generations!
Example use cases
- Fixing hands - use a mask to highlight a deformed hand, and re-generate using the default caption
- Object interaction - use a mask to highlight the interaction with the environment, and either rewrite the caption to describe a different interaction or remove the interaction from the caption
- Removing background characters / objects - use a mask to highlight the area you'd like to remove, and remove the description of the characters / background object from the caption
- Targeted replacement - if you'd like to replace a specific object or add an element in an area, draw a mask over the area, and describe the object being added explicitly (without the entire caption). For example, if you'd like to add a lamp, you might write "3D cartoon rendering style, antique oil lamp on a desk"
For any further questions, feel free to ask us directly!
