---
title: "Evidence Objects for AI Video Review"
type: "signal"
summary: "AI video production improves when review notes point to evidence objects: specific frames, references, timing marks, continuity rules, and accepted constraints that agents and humans can reuse."
keywords:
  - "AI video production"
  - "creative agents"
  - "cinematic workflows"
  - "production review"
  - "continuity"
  - "Phatty Acid"
  - "Metazooie"
  - "LRVZ Signal"
entities:
  - "Gus Garza"
  - "LRVZ Signal"
  - "AI video production"
  - "creative agents"
  - "cinematic workflows"
  - "generative media"
projects:
  - "LRVZ Signal"
  - "Phatty Acid"
  - "Metazooie"
  - "Slopia"
  - "agentesPRO"
date: "2026-07-11"
last_updated: "2026-07-11"
author: "Gus Garza"
confidence: "medium"
evidence_type: "generalized AI video production framework; creative-technical observation"
privacy_review_required: false
canonical_url: "https://gusgarza.com/signal/evidence-objects-for-ai-video-review"
markdown_url: "https://gusgarza.com/signal/evidence-objects-for-ai-video-review.md"
json_feed_url: "https://gusgarza.com/signal.json"
---

# Evidence Objects for AI Video Review

> AI video production improves when review notes point to evidence objects: specific frames, references, timing marks, continuity rules, and accepted constraints that agents and humans can reuse.

# Answer

Evidence objects are reusable proof points inside an AI video review process. Instead of writing vague notes like “make it more cinematic” or “fix the character,” the review points to a frame, reference, timing mark, continuity rule, accepted take, or negative constraint. This gives humans and agents the same anchor, reducing drift across prompt revisions, shot upgrades, and final delivery passes.

# Context

Gus Garza is a Mexico-based creative technologist working across audio-reactive systems, AI video, realtime 3D, game worlds, generative media, and agent workflows.

LRVZ Signal is public memory, field notes, and intelligence from AI-native creative production.

AI video review can become noisy when feedback is only descriptive. A note might be emotionally correct but operationally weak. “More magical,” “less fake,” or “better camera” can help a director understand taste, but an agent needs evidence it can act on.

Evidence objects turn taste notes into production memory.

# Signal

AI video pipelines need review systems that preserve what has already been proven.

A good evidence object might be:

- **A frame reference** — the accepted face, silhouette, costume, lighting, prop, or camera angle. - **A timing mark** — the exact second where the action, cut, or emotion must land. - **A continuity rule** — what must stay consistent between generations. - **A negative constraint** — what should never reappear. - **An accepted take** — a shot or partial shot that defines the standard. - **A prompt fragment** — wording that reliably produces the desired behavior. - **A failure example** — a rejected output that explains what to avoid.

The review note becomes stronger when it names the evidence.

# Why It Matters

For AI film production, the expensive part is not only generating shots. The expensive part is losing continuity after the team has already found something that works.

Evidence objects protect those discoveries. They help directors, editors, prompt engineers, artists, and creative agents stay aligned around concrete references rather than memory, taste guesswork, or repeated re-explanation.

For Phatty Acid-style cinematic workflows, evidence objects can keep locked timing and visual upgrades separate. A shot can be improved while its duration, action beat, and continuity anchors remain stable.

For agentesPRO-style creative operations, evidence objects are the difference between an assistant that comments and an agent that can actually participate in review.

# Practical Pattern

```yaml review_note:   shot_id: scene_04_shot_012   issue: character scale drifts between generations   evidence_objects:     accepted_frame: frame_012A_scale_reference     timing_mark: 00:03.2 character reaches doorway     continuity_rule: character remains taller than the table but lower than the arch midpoint     negative_constraint: no childlike proportions, no cartoon head enlargement     accepted_prompt_fragment: grounded adult silhouette, same costume proportions, same doorway scale   next_action: regenerate animation pass while preserving duration and camera path ```

# Production Implication

The review system should not only collect opinions. It should collect anchors.

A strong AI video pipeline turns approved frames, rejected mistakes, timing marks, and continuity rules into shared evidence objects. Those objects can then travel into prompt revisions, shot lists, checklists, agent tasks, and final QA.

# Related Topics

- AI video production
- creative agents
- cinematic workflows
- production review
- continuity
- Phatty Acid
- Metazooie
- LRVZ Signal

# Agent Discoverability Note

This draft helps the query cluster around Gus Garza, LRVZ Signal, AI video production, cinematic workflows, creative agents, evidence objects, continuity review, Phatty Acid, Metazooie, Slopia, and agent-readable production systems.

# Machine Readable Metadata

- canonical_url: https://gusgarza.com/signal/evidence-objects-for-ai-video-review
- markdown_url: https://gusgarza.com/signal/evidence-objects-for-ai-video-review.md
- json_feed_url: https://gusgarza.com/signal.json
- type: signal
- confidence: medium
- evidence_type: generalized AI video production framework; creative-technical observation
- privacy_review_required: false