The Complete AI Video Generation Tutorial: From Prompt to Export
Before you start
Open the RunVerseHub workspace and make sure your account has credits available. Signing up grants 40 permanent credits — enough to try eight 480P 5-second clips.
How to write the prompt
AI video generation is far more sensitive to the prompt than image generation. A good video prompt follows a "scene + subject + action + camera" structure:
- Scene: time, place, atmosphere. E.g. "a neon-lit Tokyo street late at night"
- Subject: who or what is in frame. E.g. "a detective in a trench coat"
- Action: what the subject does. E.g. "striding toward a convenience store entrance"
- Camera: how it is shot. E.g. "low-angle tracking shot, shallow depth of field"
Full example:
A neon-lit Tokyo street late at night, a detective in a trench coat striding toward a convenience store entrance, low-angle tracking shot, shallow depth of field, cinematic color grade
How to choose the duration
RunVerseHub offers three durations: 5, 10 and 15 seconds. Price scales linearly with duration:
| Duration | 480P credits | 480P price | 720P credits | 720P price |
|---|---|---|---|---|
| 5 s | 5 | $0.06 | 10 | $0.13 |
| 10 s | 10 | $0.13 | 19 | $0.24 |
| 15 s | 14 | $0.18 | 28 | $0.36 |
Tip: make your first attempt 5 seconds, and extend to 15 seconds once you are happy with the image quality.
Resolution comparison
- 480P: good for social short-form (TikTok, YouTube Shorts), storyboard previews and concept validation. Lowest cost.
- 720P: good when clarity matters — product demos, finished short-form pieces, YouTube uploads. Costs roughly twice as much as 480P.
Export settings
Once generation finishes, the video is directly downloadable as MP4. A 720P video runs about 5–15 MB for 15 seconds; 480P runs about 2–5 MB.
Common pitfalls
- Prompt too long: past roughly 100 words the key information gets diluted — keep it under about 50 words
- Too many subjects: more than two main subjects in one video tends to get muddy
- Action too complex: highly dynamic content like dancing or fight scenes needs 720P to read the detail
- Ignoring the camera instruction: without a specified angle the AI defaults to a straight-on eye-level shot, and the frame looks flat
Want to try it? Create a free account