# Most important tip I've learned so far with MiniMax-H3 — retained record

> Retained reddit record from reddit:StableDiffusion.

Canonical: https://highsignal.app/data/records/70c9a0f7fd5f8431

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:StableDiffusion)  published 2026-09-27 07:05 ingested 2026-09-28 02:39

[original source ↗](https://www.reddit.com/r/StableDiffusion/comments/1wrdhbr/most_important_tip_ive_learned_so_far_with/)

## retained text



```
Same tip applies to almost any video generation solution.. and that is to think out your scene first.

Before writing that long, complicated prompt, consider breaking it up into separate scenes.

I've spent all day tweaking a prompt today until I remembered this tactic I used back when Google Veo first launched.

Example:

Prompt I worked on today was relatively simple, static shot of 3 people sitting at a table in a bar... they notice something across the room, react and then one guy sprays his drink everywhere a few other things happen, laughter, spilled drinks, etc.

Kept having issues doing one prompt for the whole scene.. so I broke it up into two.

Scene 1 was up until the character got his drink to his lips.

Scene 2 started where scene 1 ended.

Much easier for MiniMax to figure out and you'll be less likely to have that one annoying glitch that screws up the whole scene.

This tip probably isn't too new for those that have done this for awhile, but if you're new or have taken a break from AI videos for awhile, a little reminder may save you boatloads of time.  :D 
```



document fetched 2026-09-28 00:31 · canonical https://reddit.com/r/StableDiffusion/comments/1wrdhbr/most_important_tip_ive_learned_so_far_with
