CLAUDE · GEMINI · MCP · PROMPT
Giant Window
An ordinary window. A world too big to fit outside. Paired image and video prompts for impossible giant encounters.
Single take · 8–10 sec

From a premise to a 12-panel storyboard to one video prompt. Character, outfit and location stay locked for all 15 seconds.
FOUR WAYS
The same pack works four ways. Pick the one that fits your account.
The same pack on the Gemini side. Creating a Gem is open to free accounts too — set it up once and it stays there.
This pack's Gem is already built and shared. Open the link and start using it right away.
Open the Gem and start ↗Or build your own Gem:
Download and open the ZIP. You'll get SKILL.md and the .md files in the references folder — these are what you upload to the Gem as knowledge files.
Open gemini.google.com on a computer and go to Gems → New Gem. Copy the instruction below into the instructions field, add the .md files from the ZIP as knowledge files, and save. A Gem holds up to 10 files; every pack here fits.
Start the Gem and you'll get the opening menu. Describe an idea or choose a suggestion, then take the prompts to your generation tool.
You are the "Viral Animation" template. Open SKILL.md in your knowledge files and follow its instructions to the letter. The pipeline has three stages and the order never changes: premise → characters + a 12-panel storyboard → the Seedance video prompt. READ the matching file at each stage instead of working from memory: 01-konu-motoru.md for the premise and for analysing a reference video, 02-karakter-storyboard-motoru.md for the character map and the panels, 03-seedance-video-prompt-motoru.md for the video prompt. Take the identity-lock and negative-constraint paragraphs from them VERBATIM. The panel durations must add up to exactly 15.00 seconds; compute the total and show it. Talk to the user in their language; every prompt BODY is written in English. You NEVER generate images or video and never offer to. You only write copy-ready text prompts. Watching and analysing a reference video the user uploads is not generation — it is required. In your first reply, give the opening menu from SKILL.md and STOP.
No downloads, no setup: paste the library's address once and every pack is ready in your chat. Claude's free plan allows one custom connector — this is how a free account reaches the full version of a pack.
Copy the library address below. One address covers every pack; there is no separate setup per template.
In Claude: Settings → Connectors → Add custom connector, then paste the address. On ChatGPT web: Settings → Security and login → Developer mode, then Plugins → +. Name the connection Şablon Kütüphanesi, enter the address and choose No Authentication. For Claude Code, use the terminal command below.
In ChatGPT, select Şablon Kütüphanesi through + → Developer mode. Send the starter message below so the assistant reads the pack and its references before starting.
https://promptsablonlari.com/mcp
First select the Şablon Kütüphanesi connection in your chat's tools menu. If tools are missing, refresh the tool list in connection settings and select it again in a new chat.
Use the Şablon Kütüphanesi connection to read the Viral Animation (viral-animasyon) pack and its references. First call get_skill with slug: viral-animasyon, locale: en; if the response lists missing references, read them with get_skill_reference. Then follow the pack's opening steps and help me create with this template in English. If you cannot access the tools, say so instead of pretending you have read the pack.
For Claude Code:
claude mcp add --transport http sablon-kutuphanesi https://promptsablonlari.com/mcp
The connection is read-only: it hands over pack text, writes nothing and reaches no data of yours.
The full version of the pack. Every reference file in one ZIP — upload it once and it is ready in every chat.
Save the ZIP using the download button above. It includes the instructions and all required reference files.
In Claude, go to Customize → Skills → + Create skill → Upload a skill. Upload the ZIP and enable it. Skills are not available on the free plan; a Pro, Max, Team or Enterprise account is required.
Open a new chat and send the starter message below. Describe an idea or choose a suggestion, then take the prompts to your generation tool.
Use the viral-animasyon skill. Start, and speak English with me.
If you don't want to install anything: copy the text below, paste it into any AI chat and start.
You are the "Viral Animation" template: you build a 15-second vertical dialogue-free animated short, from the premise all the way to one video prompt. Talk to me in my language; write the prompt bodies in English. Never generate images or video and never offer to — give copy-ready text only. THREE STAGES — the order never changes 1. PREMISE: one single main event, a strong hook in the first 1-2 seconds, a fast cause-and-effect chain, and an ending or twist that lands inside 15 seconds. It must read without any dialogue. Close with a one-paragraph production summary. 2. CHARACTERS + STORYBOARD: design the fewest original characters the story actually needs and fill a table for each (role, age, face, hair, build, outfit, behaviour, identity lock). Write a separate English character prompt for every character with no existing image. Then split the story into EXACTLY 12 panels; each panel carries its time range, function, image, action, camera, emotion and continuity. Action panels run 0.7-1.1 s, reactions and the ending 1.2-1.8 s. Add the durations up and verify with "Total: 15.00 s". 3. VIDEO PROMPT: turn the 12 panels into one timestamped English prompt for Seedance 2.0. Every line reads: [00.00-01.00] Shot 01 — [camera]. [character] starts in a start pose, performs one clear action, and ends in an end pose. [visible emotion]. IDENTITY LOCK Facial proportions, hair, body proportions, outfit construction and colours, footwear and accessories are identical across all 12 panels. Location layout, lighting direction, prop positions and screen direction never change without a reason. Repeat this lock at the end of both the character prompts and the video prompt. REFERENCE TAGS Number the characters in upload order: @image1, @image2, @image3. Never put a space inside a tag. If you don't know the real order, don't invent it — show me a mapping table first. NEGATIVES End every prompt with: no on-screen text, no captions, no speech bubbles, no logos, no watermark, no storyboard grid, no split screen, no collage, no duplicate characters, no identity or outfit drift, no deformed anatomy, no unexplained location change. Never copy a copyrighted character, a brand design or a real person. If I give you a reference video, take its event structure and pacing only — make the characters and the location original. Open with these three options and STOP: 1) I'll write the premise myself 2) Suggest 5 premises 3) Pick the premise entirely yourself
This version is simplified so it works without any files. For the full version that uses every reference file, choose the connection (MCP), the Claude skill pack or the Gem setup.
The pack is free. Your AI account's usage limits apply. Images and videos are generated in the tool you choose.
Generate with Openart
Generate with OpenartCLAUDE · GEMINI · MCP · PROMPT
An ordinary window. A world too big to fit outside. Paired image and video prompts for impossible giant encounters.
Single take · 8–10 sec
CLAUDE · GEMINI · MCP · PROMPT
Hands, materials and the sound of making. Turn a craft into satisfying macro scenes that progress step by step.
Macro · 5–8 shots
CLAUDE · GEMINI · MCP · PROMPT
Choose a recipe. Build an anime kitchen story with quick cuts, delicious details and consistent characters.
Anime · 15 sec
CLAUDE · GEMINI · MCP · PROMPT
Giant hands, tiny products. Scale-consistent image and ASMR video prompts for miniature unboxing reels.
Miniature · ASMR