Curated AI video prompts for Kling 3.0, Veo 3, Wan 2.5, Runway and Seedance, with model, settings and why each one works.

Quickstart · The list · How it works · Examples · Install · FAQ

English · Русский · 简体中文 · Español · Português (BR) · 日本語 · 한국어

Awesome AI Video Prompts: Kling 3.0, Veo 3, Wan, Runway, Seedance

A curated list of AI video prompts (text-to-video and image-to-video) for Kling 3.0, Veo 3, Wan 2.5, Runway Gen-4 and Seedance 2.x, plus a short set of image prompts for Flux, Nano Banana and GPT Image. Every entry carries the prompt in a code block, the model it is written for, mode, duration, aspect, a note on why the structure works, and a rendered_on field that stays empty until someone reports a real render. The data is JSON; the Markdown pages and the counts below are generated from it.

Model-specific, not generic	Honest about results	Works as an agent skill
Each prompt follows the vendor's own guide (cited with a date): one camera move for Kling, audio in its own sentence for Veo, labelled shots for Seedance.	"Written for" and "rendered on" are separate fields. The counts in this README come from scripts/stats.py, not from a keyboard.	SKILL.md lets Claude Code, Cursor or Codex search the list and adapt the closest prompt to your shot.

Quickstart

git clone https://github.com/uniqaiapp/awesome-ai-video-prompts
cd awesome-ai-video-prompts
python3 scripts/find.py --model kling-3 --use-case product

You get the matching prompts as text, each with a Why it works line and whether a render has been reported. No packages to install; Python 3.9+. Browse without the terminal: by model or by use case.

Demo

What is in the list

Model	Prompts written for it
Kling 3.0 (kling-3)	32
Veo 3 (veo-3)	7
Wan 2.5 (wan-2.5)	12
Runway Gen-4 family (runway-gen-4)	3
Seedance 2.x (seedance-2)	6
Flux (flux)	6
Nano Banana (Gemini image) (nano-banana)	8
GPT Image (gpt-image)	8

Use case	Prompts
Camera moves	9
Product	6
Portrait	5
Nature	5
Action	5
Ads	4
Anime	4
Image to video	6
Stills (image prompts)	8

Mode	Prompts
text to video (t2v)	38
image to video (i2v)	6
text to image (t2i)	6
image edit (i2i)	2

52 prompts in total: 44 video, 8 image. 7 carry a render report (rendered_on is not empty); 45 follow a vendor prompt guide or the UniqAI library but have no render report yet.

The list

One line per prompt; the link opens the full entry with the prompt text, the why-it-works note and any render reports. Model keys: kling-3 Kling 3.0, veo-3 Veo 3, wan-2.5 Wan 2.5, runway-gen-4 Runway Gen-4 family, seedance-2 Seedance 2.x, flux, nano-banana, gpt-image. Modes: t2v text to video, i2v image to video, t2i text to image, i2i image edit.

Camera moves

• Brass compass, push in · kling-3 · t2v · 5 s
• Lit candle, pull back · kling-3 · t2v · 5 s
• Harbour boats, pan left · kling-3 · t2v · 5 s
• Hiker on a ridge, tilt up · kling-3 · t2v · 5 s
• Bronze horse, orbit · kling-3 · t2v · 5 s
• Lighthouse at dusk, dolly in with sound · veo-3 · t2v · 8 s
• Old town rooftops, crane up · wan-2.5, kling-3 · t2v · 5 s
• Handheld follow through a corridor · runway-gen-4, seedance-2 · t2v · 5 s
• Chrome droplet over a neon street, push in · kling-3 · t2v · 8 s

All camera moves prompts with the full text →

Product

• Headphones on a turntable · kling-3 · t2v · 5 s
• Perfume bottle, macro water drop · kling-3 · t2v · 5 s
• Sneaker, quarter orbit · kling-3 · t2v · 5 s
• Coffee pour, slow motion with sound · veo-3 · t2v · 8 s
• Watch, two connected shots · seedance-2 · t2v · 10 s
• Skincare jar, slow rotate on marble · kling-3, wan-2.5 · t2v · 5 s

All product prompts with the full text →

以上为 README 的开头部分。其余内容请阅读原文。 查看完整 README ↗