🦥 AutoVisuals

Automated Illustration & Prompt Generation Engine

AutoVisuals designed by Sloths Intel, is a fully automated pipeline for generating randomised Midjourney-ready prompts, sending them to Discord, automatically downloading and splitting MJ images, and building a stylish HTML gallery with zoom navigation, for business design, internal datasets, Adobe Stock, and other illustration stocks.

---

Contents

• Features
• Installation
  • From pip
  • From conda
  • From source(advanced)
• Required Environment Variables
• Optional Scaling Modules
  • Install Real-ESRGAN
  • Install SwinIR
  • Install Topaz
• Usage
  • Pipeline
  • Subcommand
• Free Providers Included
• Theme List Format
• Future Modules
• Contribution
• License
• Links

---

Features

Prompt & Metadata Generator

• Generates output: theme, title, description, 45 keywords, and /imagine prompt.
• Supports output formats: txt, csv, and json.
• Supports stock metadata formats: Adobe Stock, Shutterstock, and Freepik.

Discord Automation

• Sends each prompt line to any Discord channel via webhook.
• Confirms each prompt in your private server with one click.
• Downloads MJ bot images via Discord bot token.
• Auto-splits 2×2 grids into 4 tiles.

Scaling Processor (optional)

Coming soon

HTML Gallery Builder

• Builds a techno-tidy responsive gallery:
  • Date → Category → Images
• Zoom mode includes:
  • Prev/Next navigation.
  • Back to Gallery.

Full Pipeline Command

Run autovisuals pipeline to get a pipeline of generate → send → download → split → scale(optional) → gallery.

Status Summary

Run autovisuals status to show how many prompts/images exist per date/category.

⬆️

---

Installation

From pip

First,

conda create -n autovisuals python==3.11
conda activate autovisuals
cd <your_work_directory>

Coming soon

From conda

Similarly,

conda create -n autovisuals python==3.11
conda activate autovisuals
cd <your_work_directory>

Coming soon

From source (advanced)

conda create -n autovisuals python==3.11
conda activate autovisuals

Clone the repository.

git clone https://github.com/slothsintel/AutoVisuals
cd AutoVisuals

Install environment.

pip install -r requirements.txt

Add to PATH.

echo 'export PATH="$HOME/AutoVisuals/scripts:$PATH"' >> ~/.bashrc
source ~/.bashrc

Install scaling proccessor (optional):

See details here

⬆️

---

Required Environment Variables

For prompt generation, where to get openai api:

export API_KEY="your LLM API key"

For Discord prompt sending, where to get discord webhook:

export WEBHOOK_URL="https://discord.com/api/webhooks/..."

For Discord image downloading, where to get discord bot token and mj channel id:

export DISCORD_BOT_TOKEN="your-bot-token"
export MJ_CHANNEL_ID="123456789012345678"

---

Install Real-ESRGAN

Follow this or original instruction to install Real-ESRGAN. Quick install:

cd AutoVisuals
git submodule add https://github.com/xinntao/Real-ESRGAN.git Real-ESRGAN
cd Real-ESRGAN
pip install basicsr
pip install facexlib
pip install gfpgan
pip install -r requirements.txt
python setup.py develop

Install SwinIR

Coming soon

Install Topaz

Coming soon

⬆️

---

Usage

Pipeline

Pipeline command autovisuals pipeline to generate → send → download → scale(optional) → gallery.

Option:

-h, --help        show this help message and exit.
-p, --provider    chatbot provider, choose openai by default, anthropic, gemini, llama, or deepseek.
-l, --list        list of visuals list, choose autovisuals/data/adobe_cat.csv by default or others.
-m, --mode        mode to generate prompts by themes, choose r(random) by default or m(manual).
-t, --title       title to generate prompts, choose r(weight random) by default or m(manual) input.
-d, --records     number of prompts for each run, 5 by default.
-r, --repeat      number of times to repeat each prompt for diversity, 2 by default.
-o, --out         prompt output directory, prompt/<date>/<theme> by default.
-w, --webhook     webhook URL, need to export it as environment variable.
--download-dir    images download directory, mj_downloads/<date>/<theme> by default.
--gallery-out     gallery file output directory, mj_gallery.html by default.
--idle-seconds    downloader idle timeout in seconds to proccess gallery, 120 by default.
-U, --upscale     optional upscaling step after download (y = RealESRGAN, default: n)..
--export-dir      export root for upscaled images (absolute path,
                  e.g. /mnt/c/Users/xilu/Downloads/autovisuals_export).

Subcommand

Subcommand autovisuals generate to generate prompts + metadata.

Option:

-h, --help        show this help message and exit.
-p, --provider    chatbot provider, choose openai by default, anthropic, gemini, llama, or deepseek.
-l, --list        list of visuals list, choose autovisuals/data/adobe_cat.csv by default or others.
-m, --mode        mode to generate prompts by themes, choose r(random) by default or m(manual).
-t, --title       title to generate prompts, choose r(weight random) by default or m(manual) input.
-d, --records     number of prompts for each run, 5 by default.
-r, --repeat      number of times to repeat each prompt for diversity, 2 by default.
-o, --out         prompt output directory, prompt/<date>/<theme> by default.

Subcommand autovisuals discord to send prompts to Discord webhook.

Option:

-h, --help        show this help message and exit.
-o, --out         prompt output directory, prompt/<date>/<theme> by default.
--category        specific category slug to send, true by default.
--all-categories  send prompts for all categories for latest date, true by default.

Subcommand autovisuals download to download Midjourney images.

Option:

-h, --help        show this help message and exit.
-t, --token       discord bot token, need to export it as environment variable.
-c, --channel-id  discord channel id, need to export it as environment variable.
-o, --out OUT     images download directory, mj_downloads/<date>/<theme> by default.
--limit LIMIT     stop after N images, no limit by default.
--idle-seconds    downloader idle timeout in seconds to proccess gallery, 120 by default.

Subcommand autovisuals gallery to build HTML gallery.

Option:

-h, --help        show this help message and exit.
--download-dir    images download directory, mj_downloads/<date>/<theme> by default.
--prompt-dir      prompt output directory, prompt/<date>/<theme> by default.
--out OUT         gallery file output directory, mj_gallery.html by default.

Additional command autovisuals status to show a tidy summary of prompts + images per date/theme.

Example:

DATE         CATEGORY             PROMPTS   IMAGES
2025-11-21   business                  3        12
2025-11-21   nature                    3        12
---------------------------------------------------
TOTAL                                6        24

Additional command autovisuals meta to create metadata uploads to Adobe Stock/Shutterstock/Freepik.

⬆️

---

Free Providers Included

AutoVisuals now includes two completely free API providers:

Llama (Llama 4 Maverick)

• No API key required
• High performance
• Good for bulk generation
• Endpoint: https://api.llama-api.com/chat/completions

DeepSeek (DeepSeek V3)

• No API key required
• Extremely fast
• Stable JSON outputs
• Endpoint: https://api.deepseek.com/free/chat/completions

⬆️

---

Theme List Format

Each themes and its weights are in the same row, seperated by comma.

theme,weight
forest in fog,4
business teamwork,3
sunset over mountains,5
......

---

Future Modules

• SwinIR installation
• Topaz installation
• pip installation
• conda installation
• Windows installation
• GUI (AutoVisuals Studio)

---

Contribution

Maintained by Sloths Intel GitHub, and Daddy Sloth Github.

---

License

© 2026 Sloths Intel.

A trading name of Sloths Intel Ltd Registered in England and Wales (Company No. 16907507).

MIT License.

---

Links

• AutoVisuals Website
• AutoVisuals GitHub
• Company homepage

⬆️