Command Reference

Generate an anime image from a comma-separated tag prompt. Supports optional LoRA style adapters.

• prompt (required) — Comma-separated tags. Lead with 1girl or 1boy, add visual tags, and quality boosters.
• image_name — Filename for the output (default: test).
• batch — Number of images, 1–10 (default: 1).
• seeds — Space-separated seed numbers to reproduce previous results.
• keep — Keep the file on the host machine after sending (default: false).
• auto_quality — Append masterpiece, best quality, very aesthetic, … when your prompt is missing them (default: true).
• style — Apply an art-style preset that injects the right tags for you: Anime coloring, Manga, Watercolor, Oil painting, Sketch, Cel shading, Chibi, Flat color, or Lineart.
• negative — Override the server's negative prompt for this one request.
• lora — Optional LoRA style adapter. Start typing to autocomplete from the available list.
• lora_strength — How hard the LoRA pulls, 0.0–1.5 (0.6–1.0 typical).
• lora2 — Optional second LoRA to stack on top of the first. Requires lora to be set.
• lora2_strength — Strength of the second LoRA, 0.0–1.5.

Convert an attached image into anime style.

• image (required) — Image attachment to restyle.
• prompt — Optional extra tags to guide the result.
• strength — 0.1–0.85 (default 0.7). Lower = subtler, higher = bigger restyle.
• batch — Number of styled variations to generate, 1–4 (default 1).
• image_name — Filename for the output (default: img2img).
• keep — Keep the file on the host machine after sending (default: false).
• auto_describe — Vision AI auto-tags the source (default: true).
• negative — Override the server's negative prompt for this one request.

Upscale an attached image to roughly 2× resolution using APISR, an anime-tuned super-resolution model.

• image (required) — Image attachment to upscale.
• image_name — Filename for the output (default: upscaled).

RickyBot invents a brand-new character and scene every time and draws it — no typing needed. The prompt is shown with the result, so you can reuse or tweak it.

Assemble a prompt from dropdown menus instead of typing tags. Pick a category, tick tags from the multi-select, switch categories to keep adding — your picks accumulate and are auto-ordered so the subject leads — then hit Generate. Quality boosters are added for you. The panel is private to you and times out after a few minutes.

List all installed LoRAs and their trigger words. Shows up to 25 entries; use the lora: autocomplete in /generate to browse the full list.

Browse tag suggestions across ten categories: quality boosters, art styles, lighting, composition, mood, settings, hair, eyes, clothing, and special effects.

Show pending requests, current generation status, average time per request, and a rolling ETA computed from recent jobs.

Cancel your own pending requests that haven't started yet.

Show recent completed generations, including prompts and seeds.

Talk to RickyBot's AI personality (Qwen 3.8 27B on Groq) for assistance, prompt help, or open-ended dialogue.

• message (required) — What you want to say to the AI.

Mention the bot or reply to its messages to chat without slash commands. Attach an image (up to three) and it reacts to what it sees; ask for a prompt ("make a prompt for this") and it writes a ready-to-paste /generate tag list. You can also @mention it in a reply to someone else's image post. It also joins conversations on its own now and then. On the AI chat page of the web dashboard, server admins choose where it talks, how often it chimes in on its own (0% = never), and a cap on how many times it answers one member in a time window (Administrators and chosen roles are never limited).

Memory: RickyBot remembers little things people tell it — favorite games, what someone's working on, the server's inside jokes — and brings them up later. Each server's memory stays in that server. Admins can read, edit, add, delete or switch it off on the dashboard's AI chat → Memory page. Want a fresh start? Just ask RickyBot to forget you — it wipes its notes and chat history about you in that server.

View or set personal generation defaults that apply to your requests only, layered on top of the server's settings. Run with no options to see your current effective values.

• resolution — Choose Portrait (832×1216), Landscape (1216×832), Square (1024×1024), or HD (1344×768).
• steps — Your preferred inference steps, 1–28 (28 is the art model's recommended maximum, and everyone shares one GPU).
• reset — Set to true to clear all personal overrides and return to server defaults.

Toggle whether the bot mentions you when your generation finishes. On by default — run once to silence pings, run again to re-enable.

Display in-Discord help with a quick command summary.

View dimensions, inference steps, guidance scale, sequential-batch setting, and the active negative prompt for this server.

Check GPU name and real-time VRAM usage status.

Credits, live server/member counts, the tech stack powering the bot (Animagine XL 4.0, APISR, and Qwen 3.8 27B via Groq), and quick links to the website, Terms of Service, and Privacy Policy.

Retrieve a link to invite RickyBot to another server.

Add your name and avatar to the wiki Thank You Wall. Requires an active Buy the Dev a Coffee! entitlement.

Remove your name from the Thank You Wall. You can re-opt in any time with /showme.

Warn a member. Warnings are logged and count toward automatic escalation.

• user (required) — Member to warn.
• reason (required) — Why. Sent to them by DM if DMs are on.

Time a member out so they can't chat or join voice. /untimeout lifts it early.

• user (required) — Member to time out.
• duration (required) — e.g. 10m, 2h, 1d, 1w (max 28 days).
• reason — Why.

Mute a member in every channel. Mutes use Discord's timeout, so they work however your roles and channel permissions are set up — unlike a classic mute role. Muted members also get a visible Muted role (turn it off with /automod muterole). Leave the duration empty to mute until /unmute; mutes longer than Discord's 28-day limit are renewed automatically, and leaving and rejoining doesn't escape a mute. /tempmute is the same with a required duration.

• user (required) — Member to mute.
• duration — e.g. 30m, 2h, 3d, 2w. Empty = until unmuted.
• reason — Why.

Kick a member from the server.

Ban a user — they don't have to be in the server. Add a duration for a temporary ban that lifts itself. /unban takes a user ID.

• user (required) — User to ban.
• duration — e.g. 1d or 1w. Leave empty for a permanent ban.
• delete_messages — Also delete their messages from the last hour, day, or week.
• reason — Why.

Bulk-delete 1–100 recent messages in the current channel, optionally only from one member.

Set a channel's slowmode in seconds (0 turns it off).

A user's moderation history — warnings, timeouts, kicks, bans and auto-mod hits. /clearwarns clears their active warnings.

Add emojis to the server: paste emojis from another server (or their links — right-click an emoji → Copy Link, no Nitro needed), an image link, or attach an image. Up to 10 at once; pasted emojis keep their names, and pictures over Discord's 256 KB limit are shrunk to fit (GIFs stay animated). Moderators with Manage Expressions can always use it — the dashboard's Emojis tab lets you add more roles, or switch it off. RickyBot needs Manage Expressions too.

• emojis — emojis, emoji links or image links
• image — an image to make into an emoji
• name — the emoji's name, when adding one

Pick the channel where every moderation action is logged. Leave it empty to stop logging.

Configure auto-moderation. Bad words, invite links, Discord's word presets and mention spam run as Discord AutoMod rules, so they keep working even while RickyBot is offline. Caps, emoji spam, repeated messages and external links are checked by the bot.

• show — Current setup, plus any permissions the bot is missing.
• enable — Turn moderation on or off for the server.
• rule — Turn a rule on or off and set its action (delete, warn, timeout, mute, kick, ban), threshold and timeout/mute length.
• words — Add, remove, list or clear blocked words (*word* also matches inside other words).
• links — Domains allowed past the external-links rule.
• exempt — Roles or channels auto-mod ignores. Admins are always exempt.
• escalation — Automatic punishment after N warnings within N days (timeout, mute, kick or ban).
• muterole — Turn the visible Muted role on or off.

While people chat, coins sometimes drop, and the first member to grab them gets them — by clicking the Collect button, or, if the server chooses, by typing .pick (or another word it picks) in that channel. With the optional password on, each drop's picture shows a random 1–3 character code to type after it (.pick k7) — on the picture only, so it can't be copied — a new one every drop, so a pasted .pick can't win every time. Drops follow activity: by default roughly one per 40 messages, at least 5 minutes apart, one at a time. Quiet servers get none, and one member's messages count only once every 10 seconds, so spam doesn't make more. With no drop channels picked, coins drop where people are talking.

See your coins, or another member's.

• member — optional; whose wallet to look at

Claim the daily reward (100 coins by default), once every 20 hours. Admins can change the amount or turn it off.

Send some of your coins to another member.

• member — who gets them
• amount — how many

Browse the server's shop — each item has its own Buy button, then you confirm. Admins can keep the shop to certain channels. What you buy goes to your inventory. Role items give a role when used, useable items show a message when used (and can tell the purchase-log channel, so a mod can hand something out), and collectibles are just to own. Items can be used the moment they're bought (a failed role is refunded), have a picture, custom buy and use messages, and a one-per-member limit. Roles with moderator permissions can't be sold.

See the items you (or another member) own, with a Use button on each one that can be used.

• member — optional; whose items to look at

Use an item from your inventory — a role item gives its role, a useable one shows its message. Some items are used up, others you keep.

• item — which item (start typing its name)

Give one of your items to another member.

• member — who gets it
• item — which item
• amount — how many, default 1

Put a shop item in a member's inventory for free (needs Manage Server). /remove-item takes one back.

• member, item, amount — same as /give-item

The 10 richest members of the server.

Adjust a member's wallet. Needs Manage Server.

• give member amount — add coins
• take member amount — remove coins (never below zero)

Apply a resolution preset for the server.

• preset (required) — Choose Portrait (832x1216), Landscape (1216x832), Square (1024x1024), or HD (1344x768). Sets width, height, steps, and guidance scale in one shot.

Manually tune width, height, steps, guidance, and the server's default negative prompt. Bot owner only: every server shares the owner's one GPU, and the art model is tuned for the sizes in /presets — server admins use that instead. Width/height accept 512–2048 and are snapped to the nearest multiple of 8 (an SDXL requirement) — you'll get a warning if a value was adjusted.

• width — 512–2048, snapped to a multiple of 8.
• height — 512–2048, snapped to a multiple of 8.
• inference_steps — 1–100. Higher = better quality but slower.
• guidance_scale — 1.0–20.0. How closely the model follows the prompt.
• negative_prompt — Server-wide default for what to exclude.

Revert all generation settings back to defaults.

Cancel all pending generations for a specific user.

• user (required) — The member whose queued jobs you want to cancel.

Nuclear option — clear the entire generation queue across all users.

Smoke-test the bot's three core subsystems — generation, upscale, and AI chat — and report pass/fail with timing for each. Runs at reduced settings (512×512, 8 steps) so it completes quickly without clogging the queue.

Rescan the bot's loras/ folder for newly added or removed .safetensors files. Run after dropping in a new LoRA so it shows up in /generate autocomplete.

Prune old files from the bot's outputs/ folder and report how much disk space was freed plus how much is still available. Useful when the host machine is running low.

• max_age_days — Delete outputs older than this many days (0–365, default 14). Set to 0 to wipe every file in outputs/.

Search Civitai for an SDXL-family LoRA and download the chosen file straight into loras/ — no manual hunting, no restart needed. Filtered to SFW results on SDXL, Pony, Animagine, and Illustrious base models. Picks a result from a dropdown, streams the file with live progress, then auto-refreshes the registry so it's instantly available in /generate. Returns trigger words after install when the model defines them.

• query (required) — Search terms (character name, style, artist, etc.).