JSON prompts for Veo 3: do they actually work?
The same shots written as prose and as JSON, rendered side by side, with what changed.
Alex Mercer7 min read
AI video model guides and prompts
How to prompt Kling, Veo, Seedance, Hailuo, Wan, Vidu, Runway, Luma and LTX, which model suits which shot, and what changed in each new version.
Updated 6 Nov 2026
Start here, in this order
- Which AI video model for which shot? Kling, Veo, Seedance, Hailuo, Wan, Vidu, Runway, Luma and LTXA shot-type by shot-type guide: dialogue, action, wide establishing shots, cartoon and realistic.
- Veo 3.1 prompting guide, with copy-paste templatesPrompt structure for Veo 3.1 with templates for dialogue, action and establishing shots.
- Kling 4.0 prompt guide: structure, examples, common mistakesHow Kling 4.0 reads a prompt, with examples that work and the mistakes that cost retries.
Prompt guides
Structures and templates per model.
Pixar-style AI animation prompts that stay on-model
Style words that help, style words that drift, and a template for a 3D family cast.
Alex Mercer8 min read
Version changes
What each new version changed for a series.
Kling 4.0 vs Kling 4.0 Omni: which to use for dialogue scenes
Lip movement, reference handling and price on the same dialogue scene, shot by shot.
Alex Mercer7 min read
Seedance 2.5 vs 2.5: what changed for animation
What changed between the two versions for cartoon motion, faces and price.
Alex Mercer7 min read
All verified guides in this topic shown above · Tested on Scena Studio v1.1.0
Browse all 9 blog topics →
Questions people ask
Which AI video model is best?
None for every shot: pick per shot type, then keep one model per scene for consistency.
Read the answer: Which AI video model for which shot? Kling, Veo, Seedance, Hailuo, Wan, Vidu, Runway, Luma and LTX
How should I write a Veo 3.1 prompt?
Subject and action first, then camera, then style, with one fixed line for each character.
Read the answer: Veo 3.1 prompting guide, with copy-paste templates
Do JSON prompts work better?
They help keep fields consistent between shots; the model does not read them more precisely.
Read the answer: JSON prompts for Veo 3: do they actually work?
Practical guides
How to use first and last frames in AI video
Learn when to give an AI video model both start and end images, how to plan the motion between them, and when to join shots with a cut.
How to write image-to-video prompts that control motion
Turn a still image into an AI video with a motion-focused prompt. Try three copyable examples and learn when to supply first and last frames.
About this topic
Each video model reads prompts and references in its own way, and new versions change that behaviour every few months.
These guides are written per model and per version, with templates you can paste and the shots each model handles best.
Related topics
Consistent characters
Keep faces, outfits and locations identical from shot 1 to shot 46, on every model.
Costs and APIs
What each model costs per second and per episode, and how to pay API prices instead of credits.
Comparisons and alternatives
Same shot list, same characters, different tools: comparisons with real costs.