Great frames
need direction.
From first spark to final cut: the place for mood, motion and feedback to come together.
as a feeling.
The rough cut.
Three treatments, one story. Feedback stays attached to the moment it belongs to.
Less chasing revisions. More making something worth watching.
Framewise uses AI-assisted tagging and optional draft summaries to organize notes; creative decisions remain with the team. This demo contains no actual client video or connected review service.
Framewise / Creative production
An AI-assisted preproduction workspace that translates a campaign brief into editable storyboard, shot-list and review states.
The problem
Creative teams hop between briefs, chat threads, folders and review links. AI-generated outputs can speed exploration but often obscure intent, versions and approvals.
Product strategy
Combine a cinematic timeline with visible provenance and version history. The agent proposes drafts, while people direct taste, selection and final approvals.
Proposed technical architecture
Next.js, TypeScript, Prisma, PostgreSQL, object storage, job queues and multi-agent orchestration with human approval. Synthetic briefs and generated placeholder shot descriptions. This architecture is a proposal, not a deployed backend.
Design decisions
Art-directed audiovisual / charcoal, clay and mauve. Connect original brief passages to each editable storyboard recommendation and review state.
UX research approach
Synthetic research: 6 fictional production-team interviews, 12 modeled storyboard tasks. This is a fictional scenario: no real interviews or measured outcomes are claimed.
Accessibility & responsive design
Keyboard-operable controls, visible focus, readable content and responsive reflow support this concept. Reduced-motion preferences remove animated scrolling. Physical-device and assistive-technology validation remain future work.
AI-assisted workflow
Prototype prompt-to-storyboard journeys, asset traceability, role handoff, rejected-idea recovery and review states. AI-assisted implementation is used for prototyping and iteration; product recommendations here are deterministic local simulations.
Limitations & lessons
Missing asset rights or unclear briefs flag an explicit review requirement. No automatic publishing, copyright assumptions or unapproved client-facing outputs. The next step is real user research, production data validation and accessibility testing before any launch.
Information architecture & user flow
- Creative director enters an original campaign brief
- Agent drafts beats, shots and production checklist
- Human revises or regenerates individual suggestions
- Approved assets move to a review-ready version
Core views: Brief and creative direction canvas → Editable storyboard and shot list → Version compare and sign-off queue