Pinch Her Cheeks: The Tsundere Strikes Back — AI Anime Short from Neta Studio

Pinch Her Cheeks: The Tsundere Strikes Back is a 15-second, 9:16 vertical anime short shot from an overhead first-person viewpoint in a realistic minimalist bedroom at late afternoon. You — the unseen camera wearer — reach down with a black-cuffed hand and give a playful pinch to the cheek of the silver-white-haired dragon girl lying on the bed. What follows is a perfect little tsundere meltdown: surprise, protest, a wrist catch, and a shy confession that she would not mind a gentler pinch next time.

The girl stays pure hand-drawn 2D anime — clean ink outlines and cel shading, expressive anime eyes — while the room around her stays real: pale warm-gray walls, white bedding, light oak floor and sheer curtains lit by soft window light. All five beats are voiced with a classic tsundere delivery, and there is no music — just room tone, fabric rustle and a flustered protest. A single sentence in Neta Studio grew into this fully voiced 15-second scene — world, character, motion and voice, generated end-to-end by AI.

Watch the full short: Pinch Her Cheeks: The Tsundere Strikes Back →

15 seconds · 9:16 vertical · 1080×1920 @ 24fps · voiced tsundere dialogue, no music — rendered with Seedance 2.5 (image-to-video) from the world’s own 2D character design sheet and overhead bedroom reference, with the full beat-by-beat timeline composed in the prompt.

Download the MP4 →

“15 seconds. 2D hand-drawn Japanese anime with clean ink outlines, cel-shaded color blocks, expressive anime eyes, controlled highlights, and subtle anime motion; a realistic minimalist modern bedroom, late afternoon, soft natural window light, pale warm-gray walls, clean white bedding, light oak floor, sheer curtains, restrained lived-in details, natural shadows. The character must remain pure 2D anime, never live-action, never photoreal, never realistic cosplay, never a 3D render. Reference image 1 is the 2D anime character design sheet: use it as the character identity, face, hair, horns, plush, costume, colors, and proportion anchor. Reference image 2 is the overhead minimalist bedroom and fixed first-person hand scene reference: use it as the environment, overhead layout, lighting, and hand anchor. Use both attached images as reference_image inputs. Do not use the pitch board as an input. Event goal: from the viewer's first-person perspective, a playful cheek pinch makes the proud anime dragon girl protest, catch the viewer's wrist, then secretly ask for a gentler pinch next time. Spatial blocking before movement: the silver-white-haired anime dragon girl with twin buns, pink strands, large pink eyes, iridescent dragon horns, icy wing-like accents, black choker, white-and-pink outfit, pale skirt, and small white dragon plush lies on the neatly made bed. The camera wearer is never shown; only one right hand in a simple black long-sleeve cuff appears from the lower edge. Keep the girl, bed, window light, and hand positions consistent across every shot. Camera: steep overhead first-person viewpoint, vertical 9:16 composition, stable readable framing with gentle handheld micro-motion only; use hard cuts between shots. Never show the camera wearer's face, body, or reflection. 0–2.5 seconds: steep overhead view of the same minimalist bedroom. The curtain moves slightly and the neutral mug and bedside objects remain in the background. The black-cuffed right hand enters from the lower edge and completes the approach toward the girl's face, thumb and index finger poised beside her cheek. Room tone, curtain rustle, and soft fabric sound only, no background music. Hard cut. 2.5–5 seconds: closer overhead view. She notices the fingers and shows one surprised expression; the fingers gently touch both cheeks without squeezing yet. She looks directly toward the camera and says, "等、等一下！你要做什么？" Keep enough uninterrupted time for the full line, with her mouth movement aligned to the words. Hard cut. 5–8.5 seconds: close overhead face shot. The viewer's thumb and index finger complete a gentle pinch on her right cheek, visibly compressing it softly. She shows one annoyed expression and glares toward the camera while saying, "不许捏！我才没有开心！" Keep the cheek contact visible while the full line is spoken. Hard cut. 8.5–11.5 seconds: same overhead position. Her hand completes the action of catching the black-cuffed wrist, stopping the pinch. She shows one defiant expression and says, "放开我啦！" Keep the wrist grip and the line clearly visible and synchronized. Hard cut. 11.5–15 seconds: same bedroom and warm late-afternoon light. The viewer's hand completes the release. She turns her face aside, her cheeks visibly pink, then steals one glance back toward the camera with one shy expression and says softly, "下次……轻一点。" Let the line finish, then leave half a second of quiet room tone. Sound: clear young anime girl voice with a tsundere delivery; dialogue occurs exactly in the beats where written. Use room tone, curtain rustle, cloth movement, soft cheek-pinch sound, and bed-sheet movement. No background music. Hard constraints: preserve the reference character's silver-white hair, twin buns, pink strands, large pink eyes, iridescent horns, icy wing-like accents, black choker, star pendant, white-and-pink outfit, pale skirt, and dragon plush. The room stays realistic and minimalist while the girl stays visibly 2D anime. One hand only, one girl only, no duplicate character, no extra limbs, no hand clipping, no deformed fingers, no face drift, no wardrobe changes, no unintended camera wearer face, no subtitles, no text, no logos, no watermark, no cosplay, no photoreal human, no 3D CGI, no extra people, no music.”

Model: Seedance 2.5 — image-to-video from two reference images (2D character design sheet + overhead bedroom and hand reference), 15s @ 1080×1920 (9:16 vertical).

Professional — The cloud harness supports the full workflow from world building to voiced video, all running in the cloud with no local setup. Stable access to the latest models: Claude Opus 5 · GPT Image 2 · Seedance 2.5.

One-click generation — Officially tuned skills drive every step: world building, character boards, scene references, image-to-video generation and instant publishing to a shareable link.

Community & IP-first — Adapt any world or start fresh from a single sentence. Keep your frames, prompts and sprites — remix them into new videos anytime.