ShotSense

Inspiration

At Minds Eye Creative, animation direction often begins as nuanced, conversational feedback: how a character should feel, what the audience should notice, why a pause needs to last longer, or why a pose is not communicating clearly.

The problem is that this creative intention can become fragmented as it moves between review calls, messages, spreadsheets, production trackers and retake lists. An animator may receive the requested change without fully understanding the reason behind it.

ShotSense was created to preserve the director's creative intention as feedback moves through production.

What it does

ShotSense transforms a shot reference, production context and a director's natural-language note into a clear, structured animation retake.

Using GPT-5.6, it identifies:

• The intended audience response
• The creative issue weakening the shot
• Performance and expression changes
• Posing, silhouette and staging concerns
• Timing adjustments in frames and seconds
• The production stage affected
• Priority and estimated complexity
• Possible continuity consequences
• A concise animator-facing instruction

The director remains the creative authority. Every AI-generated interpretation can be reviewed, edited and approved before it becomes a production task.

How we built it

ShotSense is being developed during OpenAI Build Week using Codex as our primary development partner and GPT-5.6 as the reasoning engine inside the product.

The project began with an early internal production-management prototype called MEC Command. During Build Week, we are designing and implementing a new focused workflow for visual shot analysis, structured retake creation, director approval and task tracking.

We are documenting the new work through Codex session history, timestamped Git commits and a dedicated build log.

Challenges we ran into

The central challenge is not simply generating feedback. It is translating subjective creative direction into useful production information without removing the human director's judgment.

We also need to combine visual and written context, produce dependable structured results, avoid a generic chatbot experience and keep the workflow quick enough to be useful during a real animation review session.

Accomplishments that we're proud of

We have turned a problem experienced repeatedly across real animation productions into a focused and testable product workflow.

Instead of attempting to replace the animation director, ShotSense is designed to protect and communicate the director's intent. It connects creative judgment with practical production management.

This section will be updated as the working Build Week prototype is completed and tested.

What we learned

Building ShotSense is forcing us to convert years of instinctive animation-direction experience into explicit production logic.

We are learning where AI can genuinely assist a creative team: not by making the final artistic decision, but by organising context, identifying ambiguity and preventing important creative reasoning from being lost.

What's next for ShotSense

Our immediate goal is to complete and test the full workflow:

director note and shot reference → GPT-5.6 interpretation → director approval → saved retake task.

After Build Week, we plan to test ShotSense inside MEC's daily production pipeline and expand it to include shot continuity, review history, production-stage tracking and collaboration between directors and animators.

Built With

• api
• codex
• gpt-5.6
• javascript
• openai