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
All product prompts with the full text →
This is the beginning of the README. Read the rest at the original. Read the full README ↗
