mcp server

ai-video-prompts

Copy-ready AI video prompts, model prompting guides (Veo, Sora 2, Kling, Seedance, Hailuo…) and agent skills for text-to-video and image-to-video.

muraterdogan0/ai-video-promptscommunitylistedunreviewed

Use tool

Set up

Runs on your computer. Pick your agent and operating system.

This project does not publish an install method we can turn into a config. Follow the README instead.

README

synced Oct 22, 2026

AI Video Prompts

A practical, copy-ready library of prompts, model guides and agent skills for generating AI video.

Everything here is written to be pasted straight into a text-to-video or image-to-video model: each prompt names the shot, the camera move, the light, the style, the length and the aspect ratio, so you get a usable clip on the first or second try instead of the tenth. The model guides explain how the major video models read a prompt differently, and the skills/ folder packages the same knowledge for AI agents (Claude Code, Claude Desktop, Cursor and any agent that loads SKILL.md files).

The prompts are model-agnostic. We test and run them in Vivideo, an AI video generator that puts many of the models below in one app, so the "try it" links point there — but the prompts work anywhere the same models are available.

Quick start: generate any prompt in Vivideo
Anatomy of a good video prompt
Prompt packs by use case
Model prompting guides
Agent skills
Finishing the clip
Further reading
Contributing
License

Quick start: generate any prompt in Vivideo

In the browser. Open the AI video generator, paste a prompt from any pack, pick a model and render. Use text to video for prompts that start from words alone, and image to video when you have a product shot, a listing photo or a character frame you want to animate. Vivideo is free to start, with no watermark and no credit card.

Two ways to work:

Manual mode — you choose the model, duration, aspect ratio and resolution for each shot. This is how the prompts in this repo are meant to be used.
Agentic chat and auto-generate — describe the whole video and Vivideo writes, plans and produces it (up to 10 minutes long), including AI avatars and AI voices.

On your phone. Vivideo runs on iOS and Android, and as a ChatGPT app.

From an AI agent or code. Vivideo publishes an MCP server, a CLI and an API. Create a key at app.vivideo.ai/account/api-keys, then add the server to Claude Desktop or Claude Code:

{
  "mcpServers": {
    "vivideo": {
      "command": "npx",
      "args": ["-y", "@vivideo/mcp"],
      "env": { "VIVIDEO_API_KEY": "vv_live_..." }
    }
  }
}

Cursor uses the same shape in ~/.cursor/mcp.json. The CLI is @vivideo/cli (npx @vivideo/cli --help). Full source for the toolchain is at egocen-vivideo/vivideo-toolchain and API reference at developers.vivideo.ai. The generate-video-with-vivideo skill teaches an agent the right call order.

Anatomy of a good video prompt

A video model is a cinematographer that only hears one paragraph. Give it the decisions a director of photography would need, in roughly this order:

Part	What to decide	Example
Subject	Who or what, with 2–3 concrete visual details	"a silver-haired fisherman in a yellow oilskin jacket"
Action	One clear main action, with a verb that moves	"hauls a dripping net over the side of a wooden boat"
Setting	Place, time of day, weather, era	"grey North Sea at dawn, light drizzle"
Camera	Shot size, angle, movement, lens	"low-angle medium shot, slow push-in, 35mm lens"
Lighting	Source, direction, quality, color	"soft overcast light, cold blue tones, warm lantern on deck"
Style	Medium and look	"photorealistic, documentary feel, subtle film grain"
Duration & format	Length and aspect ratio	"8 seconds, 16:9"
Audio	Dialogue, sound effects, ambience, music (models with native audio)	"waves slapping the hull, gulls, no music"

Rules that hold across almost every model:

One shot, one main action. A 5–10 second clip can carry one clear beat. Two subjects doing two complex things usually means neither looks right. Build longer sequences from several shots — see how to make AI videos longer than 60 seconds.
Describe what the camera sees, not what the story means. "She is nervous about the interview" gives the model nothing; "she smooths her blazer twice and glances at the door" does.
Name the camera move. Static, slow push-in, pull-back, pan, tilt, tracking, orbit, crane, handheld, drone. If you say nothing, the model guesses.
Light is the cheapest way to look expensive. Golden hour, overcast softbox, neon rim light, practical lamps — pick one and say where it comes from.
Put dialogue in quotes and name the speaker. For models with native audio, write The barista says: "Oat or whole?" and describe ambience separately. Which models generate sound is tracked in which AI video models have native audio.
Say what you do not want when the tool has a negative prompt field: "no text, no logos, no extra fingers, no warped faces, no camera shake".
Set duration and aspect ratio in the settings when the model exposes them. The prompts here end with a format line such as 9:16, 8 seconds; if your tool has separate controls, set them there and keep or drop the line. The aspect ratio calculator helps when you need exact pixel sizes, and vertical vs horizontal video covers which to choose.

The long version, with worked examples, is Vivideo's text-to-video prompt guide and how to write AI video prompts. For what actually separates prompts that work from prompts that don't, see the analysis of 40,000 AI video prompts. Unfamiliar terms are defined in the AI video glossary, and the prompt-writer reference in this repo lists camera, lens, lighting and style vocabulary.

Prompt packs by use case

Pack	What's inside	Related
Product ads	Hero shots, macro reveals, liquid splashes, launch teasers	AI ad videos
UGC ads	Phone-shot creator ads, unboxings, reactions, testimonials	AI UGC videos
Social shorts	TikTok, Reels and Shorts hooks, loops and trends	Social video
Explainer videos	Motion-graphic concepts, SaaS walkthroughs, faceless narration	AI explainer videos
Real estate	Listing tours, drone exteriors, twilight shots, staging	Real estate video
Ecommerce	Catalog motion, fashion, 360 spins, lifestyle placement	Ecommerce video
Education & training	Lessons, science visuals, onboarding, safety training	Education video
Cinematic storytelling	Short-film shots, genre looks, multi-shot sequences	AI movie generator
Talking avatars	Presenter scripts, spokesperson setups, multilingual dubbing	AI avatar generator
Music & anime	Music-video shots, loops, anime and stylized animation	AI music videos
Food & restaurants	Plating, pours, sizzle, recipe steps, menu promos	Restaurant video
Travel	Destination reels, hotel promos, aerials, itineraries	Travel video

Want more? Vivideo keeps a browsable prompt library, a gallery of example videos and ready-made video templates.

Model prompting guides

Each model reads prompts a little differently. These guides cover strengths, prompt tips and three example prompts per model. Start with how to choose an AI video model and the model strengths map if you are not sure which to pick; the index has a quick comparison.

Guide	Model page
Veo	Veo 3
Sora 2	Sora 2
Kling	Kling
Seedance	Seedance 2.5
Hailuo	Hailuo
WAN	WAN
PixVerse	PixVerse
LTX Video	LTX Video
Luma Ray	Luma Ray
Vidu	Vidu
Hunyuan Video	Hunyuan Video
Grok Video	Grok Video
Marey	Marey

All models available in Vivideo are listed on the AI models page.

Agent skills

The skills/ folder contains three skills in the standard SKILL.md format:

ai-video-prompt-writer — turns a rough idea into a production-ready prompt with model choice, aspect ratio, duration and negative cues.
generate-video-with-vivideo — generates a video through the Vivideo MCP server with the right call order, credit checks and error handling.
social-video-batch — plans and generates a batch of short-form videos with hooks, variants and per-platform aspect ratios.

Install by copying a skill folder into ~/.claude/skills/ (or your agent's skills directory). Details in skills/README.md.

Finishing the clip

Generation is rarely the last step. Free browser tools from Vivideo for the usual clean-up:

Cut and join: video trimmer, video merger, loop video, reverse video, speed controller
Reframe: video resizer, crop video, rotate video
Text and sound: subtitle generator, add text to video, add music to video, mute video, video to MP3
Files: video compressor, video converter, MOV to MP4, WebM to MP4, MP4 to GIF
Planning: script timer (how long a voiceover will run), video size calculator

The full list is on the tools page.

Contributing

New prompts, model notes and fixes are welcome. Prompts should be tested on at least one model, specific (shot, camera, light, style, format) and free of real people's names, trademarks and copyrighted characters. See CONTRIBUTING.md.