CATEGORY
Curated video generation skills, prompts, workflows and tools.
A ChatGPT Image 2.5 prompt that turns an orange-and-white clay cat — pixel sunglasses, studded collar, cigarette in mouth — into a 24-frame stop-motion skateboard loop, delivered as a finished GIF in one shot.
A free video prompt that turns a three-panel triptych image into a cinematic anime autumn park clip: the same girl on a bench, walking a leaf-covered path, then holding a red maple leaf as the camera pushes in.
A free 10-second video prompt for Gemini Flash: tiny explorers in khaki expedition gear trek across an oversized tropical island, from a low tracking shot in the jungle to a wide aerial reveal.
A free 21-second vertical stop-motion video prompt for Seedance 2 on PixVerse: a tiny humanoid figure built from beach stones dances in a tidal pool, then collapses back into ordinary stones.
A free, shot-by-shot video prompt for a photorealistic 9:16 macro clip: a miniature grey kitten no bigger than a fingertip is squeezed out of a tube onto the back of a hand and leaves a pink paw pad imprint behind.
A video prompt by Soaima for Seedance 2.0. Feed it a reference image of a glass snow globe holding a miniature McDonald's, ringed by pink, white and cream roses, and it produces a dreamy macro product clip - slow dolly push-in on the glass, falling petals, a soft breeze through the flowers, and elegant bokeh, in photorealistic 4K vertical 9:16.
A video prompt by Soaima for image-to-video models. Feed it three watercolor autumn reference images and it produces one seamless cinematic vertical clip - reaching for falling leaves, reading under the tree, riding a bicycle through the park - holding the same girl, outfit and style across all three scenes.
A detailed Seedance 2.5 prompt for a 30-second 1080p ultra-realistic early-2000s DV home video of a young Korean woman running relaxed Sunday errands in an older Seoul neighborhood. The prompt locks subject identity, scripts the trip beat by beat, and locks camera, audio, object continuity and a hard negative list so the model can't cheat its way through the shot.
A lightweight workflow for continuous multi-look outfit change videos, one character and one location, seven outfits in a single clip. Originating from the Chinese "empress styling" format, it runs on MiniMax H3 for video with prompts written by GPT-6 Astra, and the lazy version is as simple as swapping the character image and reusing an existing image-to-video prompt. Well suited to fashion and OTTD outfit content.
A reverse-engineering prompt workflow for the viral AI subway outfit-change video format. Built with GPT-6 Astra and tuned for MiniMax H3 image-to-video mode, it turns a reference video into an H3-ready prompt in one pass, so the shopping-app-to-new-outfit transformation can be recreated without an expensive big-engine video model.
A detailed 15-second video prompt that animates a cartoon character climbing out of an X profile page avatar circle, peeling the page's text off into paper planes, sweeping the leftovers away with a broom, and writing two lines across the blank space with a marker. Built for GPT-6 Astra with MiniMax H3, it takes two reference images, one for character identity and one for page layout, and locks the camera for the entire shot.
A video prompt by Soaima for Seedance 2.0. Turn an uploaded Vaseline Cocoa Glow product shot into a 10-second vertical skincare commercial in six timestamped scenes - unboxing, hands lifting the bottle, a hero shot with cocoa beans, a macro texture close-up, a rotating hand-held shot and a final push-in - while preserving the label and branding exactly.
An open-source, MIT-licensed Codex Skill that plans and scripts beat-synced outfit-change videos for a single adult female character. It locks identity from a reference image, generates five-outfit collage keyframe prompts, and outputs a shot-by-shot video timeline with 12 swappable transition mechanisms like sticker fly-ins, sleeve wipes, spin handoffs and action beats.
A short field note on Kling 3.0 for video generation. The poster compares it against Seedance and finds its handling of people noticeably more natural, eyes, smiles and body movement all land better, demonstrated with a 10-second selfie-style clip of an AI-generated woman.
No resources match these filters.
