gifs · images · how tootsies reacts

GIFs & Images

Sometimes a reaction beats words. Sometimes a picture beats a paragraph. When the moment's right, Tootsies drops a gif alongside her reply — or makes you one from scratch.

The one idea: she doesn't just grab the first result. She looks at what she's about to send, weighs whether it actually fits, and would rather say nothing than post a mid gif.
T Tootsies Today at 11:42 PM honestly? he peaked with that second album and he knows it. text reply + 🎞 reaction gif · chosen by her taste rides alongside the text reply
The gif is a bonus on top of her text reply — it never replaces what she said, it just underlines it.

Reaction gifs · how she picks

search → look → decide

It's a three-step thing. She searches Giphy for candidates, looks at the actual preview frames of what came back, and only sends one if it genuinely fits — her voice, the moment, the vibe. A corny result gets skipped entirely.

1 · SEARCH she sends a query to Giphy based on what she’s reacting to powered by Giphy 2 · LOOK she looks at the preview frames of every candidate does it read as the reaction? does it fit her voice? quick vision check 3 · DECIDE ✓ sharp → send it — or — ✗ mid → skip it she biases toward nothing if none of the candidates fit, she sends no gif — and you’d never know she tried
The search finds candidates, the vision check looks at them, and the bias is always toward “skip” over “settle.”

Generated images · two modes

generate · or remix

Ask her to draw something and she makes a fresh image. Reply to an existing image and ask her to change it, and she edits that one instead. Both fall back gracefully — if generation fails, her caption comes through as a plain text reply.

✨ GENERATE “draw me a cat dj” ↓ 🎞 brand-new image from a text prompt no source image needed ❝draw me❞ / ❝make me a❞ → generate ✏ REMIX reply to an image + ask to change it ↓ 🖼 → ✨ edits the source image reply to their pic → remix
❝draw me X❞ makes something new. Replying to someone’s image and asking to change it edits that one. Powered by OpenAI’s image model.

Pacing · so it doesn’t spam the room

cooldowns · daily caps

Neither gifs nor images are free-for-all. Gifs have a per-channel cooldown (so you won't see back-to-back gif replies) and a daily cap across the whole server. Images have a daily budget too. Both limits are checked before anything is fetched, so a capped reply costs nothing.

🎞 Gifs per-channel cooldown — waits before the next gif in that channel daily server cap — hard ceiling across all channels 🎞 Images daily server budget — caps both new pics and remixes checked before generating — a capped image sends the caption instead
Limits are mod-tunable on /menu. Defaults are generous but not infinite — the room stays clean.

The candidate funnel · many in, one out (or none)

candidates → vision check → one or skip

Giphy hands her several results. She looks at all of them — preview frames plus the context of what she’s replying to — and picks at most one. The bar is: does it read as the reaction, does it match the moment’s tone, and does it sound like her (sharp, a little ironic, not corny)? If nothing clears all three, the whole lot gets skipped. You see the winner or nothing.

CANDIDATES 🎞 🎞 🎞 🎞 from Giphy search VISION CHECK reads as the reaction? fits the moment’s tone? sounds like her? Haiku vision call ✓ ONE WINNER the one that cleared all three — posted alongside the reply ✗ SKIP (none fit) no gif sent — her text reply lands clean on its own
She’d rather send nothing than something mid. The funnel collapses four candidates to one pick — or zero.

Which mode triggers · draw vs remix

no source image → generate · source image present → remix

The decision is automatic: if there’s no image to work with, “draw me X” makes something from scratch. If she can see a source image — either in the message you replied to or attached to your mention — “make it more cinematic” edits that one. She can’t remix something she doesn’t have, so the remix tag is only offered when a source is actually there.

@mention with image ask “draw me X” / “make it Y” no source image image in context ✨ GENERATE (fresh) brand-new image from your text prompt no source needed; OpenAI makes it ✏ REMIX (edit) edits the source image you pointed at reply to an image → she edits that one
The mode is inferred from context, not a flag you set. Point her at an image and she edits it; give her nothing and she creates.

Always falls back to words

gif miss · image fail · budget spent → text reply

Nothing leaves you hanging. If a gif can’t be found or doesn’t pass the taste check, her text reply lands clean. If image generation fails, the image budget is spent, or the feature isn’t set up, the caption she wrote comes through as a plain text reply. The visual is a bonus — her answer is always the floor.

VISUAL ATTEMPT gif → search + vision pick image → generate or remix runs first, before the reply ok? yes no ✓ visual + text reply gif or image lands alongside her words text reply only her answer still goes through, always
A capped budget, a failed generation, or a “none fit” gif result all collapse to the same thing: her text reply, delivered clean.

How to use it

Just @mention Tootsie's like you normally would. She decides whether a gif fits based on the moment. For images, say it explicitly — “draw me” / “make me a” triggers a fresh generation; replying to an image and asking her to change it triggers a remix.

triggers a gif

  • any @mention she finds funny or reactive
  • posting something she'd ironically react to
  • when gifs are on and her taste approves

triggers an image

  • @mention: "draw me a sunset skyline"
  • reply to image: "make it neon"
  • when images are on and budget remains
Both are opt-in for the server. Gifs need Giphy turned on; images need an OpenAI key. If either isn’t set up, she simply won’t use that surface — and she won’t tell you she tried.

under the hoodGiphy · OpenAI GPT Image (gpt-image-2) for draw + remix · Claude Haiku (vision pick)

@mention her and ask for a pic, or just give her something to react to