MemeBot

Use in the first conversation with a new owner of this meme-maker bot, to learn their taste and connect what it needs before making memes.

Use first whenever someone asks for a meme (image, video, dub, or edit). It runs the loop from joke to real template, compose, visual check, delivery, versioned iteration and logging, and says which other meme skill to load at each step.

Use when you need the real base image or clip for a meme, such as a canonical blank template, an exact show or movie frame, a high-res episode screencap, a YouTube clip, or a real photo of a person. Covers where to look, how to download, how to verify it's the right scene, and how to fix low resolution.

Use when putting text or graphics on a meme still with Python/Pillow. Covers classic white-outlined top/bottom captions, auto-fit, labeled callout tags with pointer lines, logo or mascot badges on faces and objects, and stacking multi-panel layouts, with overlap checks and code that has run.

Use when a meme needs a custom edit of a real photo or frame that captioning can't do, such as transforming a person (into an animal, older or younger), adding or removing objects, retouching, or changing a pose. Covers model choice, the gateway calls, prompt patterns, costs and failure modes.

Use when making a video meme from real footage. Covers grabbing clips, cutting and reordering segments cleanly, matching loudness, burning in subtitles or meme captions, tracking logos and labels onto faces, freeze frames, silent versions, and frame-by-frame verification.

Use when a meme needs motion that doesn't exist in real footage, such as animating a transformation between two stills or bringing a photo to life. Covers model choice (Kling 3.0 I2V with start and end frames first), gateway calls, prompts, costs, and checking the output.

Use when a video meme needs characters to say new words in audio (parody dubs), not just new subtitles. Covers TTS through the Vercel AI Gateway, stem separation, word-level splicing, take selection by speaker similarity, matching timing, EQ and loudness, room tone, QA, and honest limits on voice likeness.

Use when a meme or meme video is finished and must be shown to the user, handed to another agent, emailed, uploaded to the user's cloud library folder, or logged. Also use when setting up or maintaining the local template and meme library.