AI Video Prompt Guide: Write Briefs That Stick

Weak prompts read like search queries. Strong prompts read like shot lists. This guide is the Tidra default for Studio.

The five-part brief

Stack these in one paragraph (or short bullets):

1. Subject — who/what is on screen, wardrobe, age, emotion.
2. Camera — lens feel, move (dolly, orbit, handheld), framing.
3. Light & grade — practicals, time of day, color temperature.
4. Motion & timing — what changes across the take; pace.
5. Exclusions — watermarks, text overlays, extra limbs, brand logos.

Example:

A slow dolly toward a coastal lighthouse at dusk, warm tungsten in the keeper’s window, rain on glass, gentle ocean spray. No text, no watermark, no people in frame.

Text-to-video vs image-to-video

| Mode | Use when | | --- | --- | | Text-to-video | You want full freedom on composition and first frame. | | Image-to-video | Brand asset or character sheet must match frame one. |

For i2v, keep the prompt about motion and camera, not re-describing every pixel the image already shows.

Tier pick without overthinking

• Fast — explore compositions and wording.
• Balanced — most social and product drafts.
• Premium — hero shots and client-facing finals.

Details live on pricing. Model-family pages (Seedance 2.5, MiniMax H3) explain capability differences when you care about longer takes vs multimodal direction.

Iterate like an editor

Change one variable per take: camera move, then light, then exclusions. Save the prompt that worked; branch variants instead of rewriting from scratch.

Ready? Open Studio and run three Fast takes before you spend Premium credits.