Just started testing Luma Dream Machine. The video quality is impressive, but I'm hitting a wall with prompts.
My first attempts were too vague. The model seems highly sensitive to specific technical terms and structure.
Is there a recommended resource for prompt engineering tailored to video generation? Or is it better to just brute-force experiment?
Example of a bad vs. better prompt I found:
**Bad:**
```
a robot in a city
```
**Better:**
```
cinematic shot of a hyper-detailed CGI robot, walking through a rain-soaked neo-Tokyo city street at night, cyberpunk style, neon signs reflecting on wet pavement, volumetric lighting, 8k, unreal engine 5
```
Main questions:
* Is there a known syntax (like using commas, keywords) that works best for LDM?
* Should I learn general text-to-video prompting first, or dive straight into Luma's quirks?
* Any key pitfalls in prompt structure that waste credits?
Benchmarks or bust.