Inspiration## Inspiration

Screenplays often contain details that sound believable but are historically, geographically, or technically impossible. Writers and production teams may spend hours checking these details manually, while a language model can produce confident answers without showing its evidence.

We built ScriptLens AI to make screenplay fact-checking faster, more transparent, and more useful for production.

What it does

ScriptLens AI accepts a screenplay scene and extracts claims that require external verification. It searches the live web for evidence, evaluates each claim, assigns a verdict and severity, and generates a production-safe rewrite.

In our demo, a scene set in Tokyo in 1987 includes a touchscreen smartphone, Google Maps, live GPS sharing, and a direct bullet train to Sapporo. ScriptLens identifies four issues, including two critical findings, explains the evidence, and suggests historically credible alternatives that preserve the scene’s dramatic purpose.

How we built it

The workflow combines:

  • Gemini for claim extraction and editorial synthesis
  • Parallel Search API for live web research
  • Google ADK for the agent structure
  • FastAPI for the backend API
  • Google Cloud Run for hosting
  • Google Secret Manager for runtime API keys

The pipeline is:

Screenplay
→ Gemini extracts claims
→ Parallel Search gathers evidence
→ Gemini evaluates the evidence
→ Structured continuity report

Searches run concurrently, and Pydantic models keep the claims, evidence bundles, findings, and rewrites structured and inspectable.

Challenges

The main challenge was handling incomplete or conflicting web evidence. ScriptLens uses an explicit uncertain verdict, source-scoped URLs, structured outputs, and retry handling for temporary model errors.

Impact

ScriptLens can help screenwriters, producers, and continuity teams catch costly factual mistakes before filming. It turns manual research into an evidence-backed editorial workflow.

What's next

Future versions could support PDF screenplay uploads, persistent project reports, team review, scene-to-scene continuity tracking, and managed Agent Platform deployment.

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for ScriptLens AI

Share this project:

Updates

Submission history