FrameFlow

FrameFlow turns director notes into secure, AI-assisted post-production workflows with role-based collaboration, quality control, and private evidence.

Comment

Inspiration

Film post-production is highly collaborative, but creative feedback often becomes fragmented across chats, spreadsheets, and disconnected tools. A director note can lose context while moving from a producer to a supervisor, then to a VFX, sound, color, or editorial artist.

We built FrameFlow for the Google Cloud Agentic Cinema Hackathon, within the IBM Bob track context, to turn that fragmented process into a secure and traceable workflow.

What it does

FrameFlow is a multi-role post-production management system that transforms director notes into structured tasks and guides them from initial review to final quality control.

A producer creates productions, invites team members, and tracks progress. A supervisor submits director notes for Gemini-assisted analysis, reviews the generated task, selects its department and priority, and assigns it to the appropriate artist.

Artists receive only the tasks assigned to their department. They can start work, add an optional delivery link or visual evidence, and submit the task for quality control. The supervisor can then approve the delivery or return it with feedback. Every decision remains available in a production-specific history.

How we built it

The frontend is built with Next.js, React, TypeScript, and Tailwind CSS and is deployed on Vercel.

The backend is a FastAPI service deployed on Google Cloud Run. It uses:

  • Gemini on Vertex AI to classify director notes, recommend a post-production department, and prioritize work.
  • Firebase Authentication with Google Sign-In to verify user identity.
  • Cloud Firestore to persist productions, memberships, invitations, tasks, notifications, and activity history.
  • Google Cloud Storage to store optional visual evidence privately.
  • A dedicated Cloud Run service account and authenticated API requests to protect production data.

Roles and workflow

  • Producer: Creates productions, invites people, manages roles, removes members when needed, and monitors production metrics.
  • Supervisor: Reviews AI-assisted director notes, approves or rejects requests, assigns artists, and performs final quality control.
  • Artist: Works only on tasks assigned to their department, submits optional evidence, and sends work to quality control.

Membership invitations must be accepted before a user can access a production. Every task, team, and history view is scoped to the active production, keeping independent projects separated.

Challenges we ran into

The biggest challenge was not only integrating AI classification, but designing a workflow that keeps the human decision maker in control. Gemini assists with classification, but it never replaces the supervisor's approval.

We also focused on data isolation. It was important that users could belong to multiple productions with different roles while never seeing tickets, members, or history from another production. Another challenge was handling the handoff between departments while preserving feedback, evidence, and a complete activity trail.

Accomplishments that we're proud of

  • Built a complete end-to-end workflow from a director note to a reviewed final delivery.
  • Added production-specific roles, invitations, and member management.
  • Integrated Gemini through Vertex AI without exposing API keys in the client.
  • Protected backend routes with Firebase ID tokens.
  • Added private evidence uploads and authenticated evidence access.
  • Created separate dashboards for decision review, artist work, quality control, and history.
  • Deployed the full application with Vercel and Google Cloud Run.

What we learned

This project showed us that an effective creative AI workflow is not just about generating an answer. It needs clear ownership, secure access, traceability, and human approval at every important transition.

We also learned how Firebase Authentication, Cloud Run, Firestore, Cloud Storage, and Vertex AI can work together to support a practical production application.

What's next for FrameFlow

Next, we plan to add richer production analytics, due dates, notifications by email, support for more file formats, better search and filtering, and collaboration tools for larger studios.

Links

Built With

Share this project:

Updates

Submission history