Lock Brand Consistency with Veo 3.1 JSON Prompts Guide

Struggling with characters that morph unpredictably or visuals that drift off-brand in your AI-generated videos? As of November 2025, Google Veo 3.1—released on October 15, 2025—offers director-level control through structured JSON prompts. This guide explores the ‘continuity’ and ‘shot’ objects in Veo 3.1 JSON prompting, enabling precise character identity, camera movements, and repeatable workflows for professional campaigns. Whether using Vertex AI, Gemini API, or Flow, master these to eliminate inconsistencies and scale on-brand video production efficiently.

Veo 3.1 JSON prompting fundamentals

Veo 3.1, Google’s state-of-the-art video model available via Gemini API and Vertex AI, generates 720p/1080p clips up to 8 seconds with native audio. Unlike plain text prompts prone to drift, JSON-structured prompts—popularized in community guides and aligned with official prompting frameworks—provide modularity. Official docs emphasize cinematography + subject + action + context + style formulas, but JSON elevates this for production.

Key structure from pro templates (e.g., developer.tenten.co, Oct 19, 2025):

{
  "version": "veo-3.1",
  "global_style": { ... },
  "continuity": { ... },
  "scenes": [ ... ]
}

This format front-loads intent, unifies style, and locks elements across shots. Use in tools supporting structured input or as meta-prompts for Gemini to refine.

The continuity object: Locking character identity

Character inconsistency plagues AI video: faces shift subtly across clips. Veo 3.1’s ‘ingredients to video’ (reference images) shines here, formalized in JSON’s ‘continuity’ block. Define characters, props, and lighting once for reuse.

Example:

"continuity": {
  "characters": [
    {
      "id": "brand_spokesperson",
      "description": "30s confident professional woman, short black hair, blue blazer, warm smile",
      "reference_images": ["front.jpg", "profile.jpg"]
    }
  ],
  "props": ["logo_cube: matte silver, no text"],
  "lighting": "consistent soft key from screen-left, 5600K"
}

Generate references with Imagen or Nano Banana first. This boosts identity stability, per Google Cloud Blog (Oct 16, 2025). Reference up to 3 images via API for shots with preserved appearance.

Before (plain prompt drift) vs. after (JSON continuity lock) character consistency

Pro tip: Pair with Veo 3.1 Fast for iteration speed.

The shot object: Director-level camera control

Veo’s prompt adherence excels with explicit cinematography. The ‘shot’ object per scene dictates type, framing, and motion, preventing random reframing.

Example scene:

{
  "id": "s1",
  "duration_sec": 4,
  "shot": {
    "type": "medium shot",
    "framing": "shoulders-up, centered",
    "camera": "slow dolly-in, subtle parallax"
  },
  "action": "spokesperson introduces product"
}

Terms like "crane shot," "shallow DOF," "tracking" yield precise results (Vertex AI prompt guide, Nov 2025). Atomic scenes (one action) ensure adherence.

Plain Prompt Issue	JSON Shot Fix
Random zoom/crop	"framing: full subject visible"
Unwanted motion	"camera: locked-off"
Off-style	"type: macro insert"

Repeatable workflow for brand campaigns

Scale consistency: Template JSON per brand, generate refs, batch scenes.

Define globals: colors (#007BFF primary), mood (professional tech).
Lock continuity with brand assets.
Script scenes atomically.
Generate via API/Flow, extend clips.
Compile in editor.

Veo 3.1 JSON workflow for on-brand video production

API call (Python, Gemini API):

operation = client.models.generate_videos(
    model="veo-3.1-generate-preview",
    prompt=json.dumps(template),  # Structured JSON
    config=types.GenerateVideosConfig(reference_images=refs)
)

Advanced tips and real-world examples

Negative prompts: "blurry, deformed faces, text overlays." Timestamp scenes: "[00:00-00:04] action…" for multi-beat clips. Audio: "SFX: product click, ambient office hum."

"Continuity block stabilizes identity across shots (stronger in 3.1)."
developer.tenten.co guide, Oct 2025

Campaign example: Tech product launch—host consistent across 5 scenes, branded lighting, precise hero shots. Yields 30-40% retention uplift (Pocket FM case).

Conclusion

Key takeaways: (1) Use ‘continuity’ for locked identities via refs; (2) ‘shot’ objects for camera precision; (3) Atomic scenes in templated JSON for repeatability; (4) Workflow scales campaigns; (5) Iterate with Veo 3.1 Fast. Start by adapting the template in Google AI Studio or Vertex AI (preview pricing: check ai.google.dev/pricing). As Veo evolves, JSON prompting future-proofs your AI video pipeline for flawless brand consistency.

Try Veo 3.1 in Gemini API

How to Lock Brand Consistency with Veo 3.1 JSON Prompts