Every command, prefix, and trick the bot knows.
This is a private bot — commands only work for approved members. If it tells you it's private, ask Bill to add you. Anyone in a group can still set off saved trigger pictures; that part needs no approval.
/gif <search>Posts a random GIF matching your search.
/gif excited golden retrieverIn groups, /gif@naertnetworksGIFBOT <search> always reaches the bot even when a bare /gif doesn't.
/ask <question>Ask anything. The bot decides on its own when the question needs current
information, searches the web, and cites its sources as [1]-style
references with the links listed under the answer.
Works as a reply: reply a bare /ask to any text
message and that message becomes the question. Type a question alongside it and
your question wins.
Photos too: reply /ask <question> to any photo, or send a
photo captioned /ask <question>, and it looks at the picture.
A bare /ask on a photo just describes it.
Ask about the bot itself: questions like /ask how do I make a
video with sound?, /ask what can you do?, or /ask which
prefix gives me the fastest image? are answered from this very guide — the
exact command to type, with an example — instead of a guess. Add bot:
(or help:) to force that when the bot mistakes a bot question for a
general one. It only knows what is on this page.
What changed lately: /ask what's new? answers from the
changelog, newest first. changelog: (or
whatsnew:) forces that and lets you narrow it: /ask changelog: anything
about videos?. A bare /ask changelog: lists the latest changes.
| Prefix | Effect |
|---|---|
bot: / help: | Answer from this command guide (questions about the bot itself) |
changelog: / whatsnew: | What changed recently, from the changelog (/ask what's new? works without the prefix) |
meme: / funny: / joke: | Joke delivery — still answers correctly first |
serious: / srs: / normal: | Force the straight answer (the default) |
web: / search: | Force a web search |
nosearch: / offline: | Skip the web entirely |
pics: / img: | Attach real web images to the answer (never generated ones) |
nopics: | Never attach images |
local: | Answer with the backup AI — never refuses anything |
gpt: / chatgpt: | Force the main AI (the default) |
Prefixes combine in any order:
/ask what won the 2026 World Cup?
/ask meme: web: is cereal a soup?
/ask pics: what does a quokka look like?
/ask how do I save a trigger picture?
/ask what's new?/chat <message>Talk with the AI and it remembers the conversation, so follow-ups like
"make that shorter" or "what did I just say?" work. /ask is still the
one-shot question with no memory.
One conversation per chat: everyone in a group shares the same conversation, and the AI knows who said what. A private chat with the bot has its own.
Keep going without a command: reply to any of its chat answers and it continues. In a private chat with the bot, just type.
Photos too: reply /chat <message> to a photo, or send one
captioned /chat <message>.
Look something up: start the message with web: and it
searches first and lists its sources.
Start over: /newchat clears the chat's conversation. One
left alone for a few hours starts over by itself.
/chat help me plan a taco night for six
/chat web: what's the weather doing in Detroit this weekend?
/newchat/imagegen <prompt> · alias /igRenders an image from your prompt. Pick a model by starting the prompt with its name and a colon — the names in brackets are aliases and work the same. No prefix uses the bot's current default (Gemini, as of Aug 2026).
| Prefix | Model |
|---|---|
gemini: (nb:, google:) | Gemini — best quality cloud |
gpt: (chatgpt:, openai:) | ChatGPT — great at text in the image cloud |
flux2: (f2:) | FLUX.2 dev — gets lazy one-line prompts like the clouds do, slower local |
qwen: (qwen2:, qi:) | Qwen-Image — best local adherence & in-image text local |
flash: (nvidia:, qf:) | Qwen-Flash — NVIDIA-distilled Qwen local |
flux: | FLUX — fast, sharp text local |
dev: (fluxdev:, fd:) | FLUX.1 dev — best local quality local |
krea: | Krea 2 — aesthetic looks local |
zimage: (z:) | Z-Image — fastest local |
Local models are free and never refuse a prompt. Cloud models look great but can decline edgy prompts — when one does, the bot automatically re-renders on a local model and says so in the caption. Local renders also self-check: a vision model compares the result against your prompt and re-renders up to 3× if it missed (the caption notes it).
Add a style after the model prefix (or on its own). Send a bare
/imagegen in chat for the same list.
deepfried: wojak: reaction: oil: 3d: photo: epic: anime:
comic: pixel: vaporwave: sticker: plush: statue: claymation:
lowpoly: actionfigure: crayon: noir: polaroid: cursed: plain:Qwen style LoRAs (real style transfer, qwen: only, a bit slower):
anime: realism: pixel:
/ig a corgi holding a sign that says HOMELAB HERO
/ig gpt: a neon cyberpunk city at dusk
/ig deepfried: epic: the office thermostat war
/ig qwen: anime: a knight defending a server rack/edit <instruction> · alias /ieChanges only what you ask and keeps the rest of the photo. Two ways to hand
it the photo: send a photo with /edit <change> as the caption, or
reply /edit <change> to any photo already in the chat.
Still photos only — GIFs can't be edited.
| Prefix | Editor |
|---|---|
gemini: (nb:, google:) | Gemini — best quality cloud |
gpt: (chatgpt:, openai:) | ChatGPT cloud |
qwen: (local:) | Qwen editor — free, never refuses, ~8s local |
/edit give him sunglasses
/edit qwen: make it snow
/edit gpt: put a red party hat on the dog/enhance [steering]Up-reses a photo: it comes back re-rendered at a much higher resolution,
sharper and cleaner, with the content kept exactly the same. Same two ways to
hand it the photo: send one captioned /enhance, or
reply /enhance to any photo in the chat. The result arrives as a
file so Telegram doesn't recompress it — tap it to see the full
resolution. Extra words steer it. Gemini does the work; if it fails, ChatGPT
takes over automatically.
/enhance
/enhance focus on the face/merge [instruction]Combines two photos into one image. Reply to the first photo and
attach the second with /merge <how to combine them> as its caption —
or send both photos as one album with /merge as the album caption.
The instruction is optional (default: merge naturally). Cloud editors only —
gemini: / gpt: prefixes pick one.
/merge put the cat from image 2 on the couch in image 1/gifgen <prompt> · alias /ggGenerates a ~5 second silent GIF from your prompt. Takes a minute or two.
Your prompt is expanded into detailed motion beats first —
prefix raw: to render exactly what you typed.
You approve the expansion new:
before the render starts, the bot shows you what your idea was expanded into,
with buttons to ✅ Render it, use ↩️ My text
(exactly what you typed), or ❌ Cancel — or reply to that
message with your own version and it renders that instead. Only the person who
asked can decide, and doing nothing just renders the expansion after
15 seconds. Applies to /gifgen, /vidgen, and
/vanimate; raw: skips it entirely.
/animate [what should happen]Brings a photo to life: reply /animate to any photo, or send one
captioned /animate. Bare /animate adds natural motion. Stills only.
/vidgen <prompt> · alias /vg sound onLike /gifgen but a real video with sound — voices, sound
effects, and music are generated together with the picture, so describe what you
want to hear too. Takes a couple of minutes. raw: skips the
prompt expansion here as well.
/vanimate [what should happen] · alias /va sound on/animate's sibling with sound: reply /vanimate to a photo or send
one captioned /vanimate. Bare form adds natural motion and fitting sounds.
/swap [steering] newPuts a photo's character into a video: the character from your photo performs
the motion from the video. Reply to the video with a photo captioned
/swap (or reply to a photo with a video captioned /swap) — GIFs
work as the video too. Words after /swap steer the look. Videos up to
~20 MB; long clips are trimmed to the first few seconds. A real video keeps
its own soundtrack.
/musicgen <song idea> · aliases /mg, /music newComposes a full song with vocals from a one-line idea — about a
minute of music by default. Name a genre, mood, or specific lines and they make
it into the song; your idea is turned into a full music description and lyrics
first. Prefix a length like 120s: for a longer song (up to 300s — it may
end earlier on its own), and raw: to use exactly what you typed as the
music description (no lyrics). Takes a few minutes; one song or video at a
time per person.
/gg a penguin in a top hat jumping over a fire hydrant
/vg a jazz band of raccoons playing in a subway station
/va the statue slowly turns its head and hums
/mg an over-dramatic power ballad about losing the TV remote
/mg 180s: a country song where a man loses his truck, wife, and dogSave a picture or GIF under a word, and from then on a message that is
just that word posts it back to the chat. Punctuation, emoji, and
case don't matter (FRFR!! fires) — any other word in the message kills it
(my ac is broken stays quiet). It fires every time the word is said —
no cooldown. Anyone can set them off; saving and managing them needs approval.
| Command | What it does |
|---|---|
/addimage <word> (/addpic) | Save it: send a photo/GIF captioned /addimage <word>, or reply that to one. The bot files 5 tags of what it sees (well-known cartoon and meme characters by name, like pepe); the caption on every fire shows the word + tags. |
/images [word or tag] | List what's saved, with tags, who added it, and fire counts. 🔇 marks entries muted in the chat you ask from. |
/findimage <words> (/fi, /searchimage) | Search the library and get the matches back as pictures, each captioned with the word that fires it. Matches trigger words and tags; with several words, every word has to match ("angry cat"). The best 3 are posted, the rest are named in text. Doesn't count as a fire and works on muted triggers (flagged 🔇). |
/delimage <word> | Remove one — everywhere, for every chat. |
/mutetrigger <word> | Stop that trigger firing in this chat only — every other chat keeps it. /mutetrigger all silences every trigger here. Bare /mutetrigger shows what's muted here. |
/unmutetrigger <word> | Turn it back on here. /unmutetrigger all lifts everything, including a /mutetrigger all. |
/addimage frfr (as the caption of a photo or GIF)
/images dog (list saved pictures by word or tag)
/findimage angry cat (search, and get the pictures themselves back)
/mutetrigger frfr (this chat is tired of it; the others keep it)
/unmutetrigger all (bring them all back)Both work by US zip code, and the bot remembers the last zip each person used — type it once and you're done.
/weather <zip>Current conditions plus the next few days for that zip. After the first
time, a bare /weather reuses your last zip.
/lunch <zip> [anything]Finds lunch spots near a zip — real local places, not the same chains a lazy search returns. The parameters can be as specific or as vague as you like, or left off entirely:
/lunch 48226 cheap tacos
/lunch 48226 somewhere quiet with a patio, nothing fried
/lunch sushi (zip remembered from last time)
/lunchRecommendations are mined from across the web, Yelp included, and every one is backed by a cited source link — the bot won't invent a restaurant it can't point to.
/memeit [genre]Groups only: picks the funniest moment said since the last /memeit and
turns it into a short comedy clip with sound, starring the people who said it.
Each run rolls a random genre; name one to pick it, e.g. /memeit sitcom.
Takes a couple of minutes.
Genres: sitcom, nature doc, security cam, telenovela, silent film, news report, infomercial, reality tv, claymation, anime, found footage, action movie, video game, noir, western, cooking show, sports broadcast.
/jokeTells a joke. /joke news makes one up about today's headlines.
/request <idea> (/req)Wish the bot did something it doesn't? File it. Requests go on a list the
maintainer gets as a prioritized digest, so the best ideas rise to the top.
/requests shows what you've already filed and what's still open.
/ascii <name> and /lennyPosts an emoticon. Names:
lenny shrug tableflip unflip disapproval happy bear
angry cool cry confused magic wink flexType @naertnetworksGIFBOT <search> in any chat and pick a GIF
from the results — no need to add the bot to that chat. Works best from phones.