PRISM Overview

Version: 0.1.28 | Status: Beta | License: Proprietary (talw.media LLC)

What is PRISM?

PRISM (Production Resource & Information System for Media) is a unified creation and production management platform built for filmmakers. It combines screenplay editing, a structured Hollywood workflow pipeline, AI-powered creative tools, collaborative annotation, and on-set device integration into a single system that follows a project from first concept through final delivery.

PRISM is not an AI wrapper. Strip away every AI feature and it still stands on its own as a professional production system with Fountain screenplay editing, cascade permissions, department hierarchy, version control, annotation overlays, and workflow stages. The AI is additive -- it enters the same review, annotation, and approval pipeline as human-created content. The creator stays in the driver's seat.

The project traces its origins to a concept document written in March 2015 describing a screenplay-centric production system with metadata overlays, department-based permissions, and live on-set integration. AI was the missing piece at the time. Now that capable models exist, PRISM is the system that was always meant to harness them in service to the creative humans who actually make films.

Who Is PRISM For?

PRISM serves working filmmakers, screenwriters, and production teams -- particularly those working on short films, web series, and indie features. It is designed as a professional tool for people with craft and intention, not a consumer gimmick. Every feature serves the filmmaking process.

Typical users include:

• Screenwriters developing scripts through the Fountain format with AI-assisted enhancement
• Directors and producers managing productions from development through delivery
• Department heads (Camera, Art, Sound, Editorial, AD, Script, Producing, Direction) coordinating within a structured hierarchy
• On-set crews using the Nexus device for live capture, markers, and session tracking

Core Capabilities

Screenplay Editing

PRISM uses the industry-standard Fountain markup format for screenplays. The editor provides real-time preview with proper screenplay formatting, line numbers, and timing estimates. Standard Fountain is extended with # ACT sections for structural organization and .TITLE CARD: forced headings. Scripts support database-backed Git-style version control with named checkpoints and, after script lock, Hollywood revision colors (White, Blue, Pink, Yellow, Green, Goldenrod, Buff, Salmon, Cherry, Tan).

World-Building (Characters & Locations)

PRISM's unique power starts before the screenplay. The Characters page lets you design characters visually (AI-generated portraits), give them voices (AI voice generation), chat with them to discover their behavior, and fill rich profiles (want, need, false belief, backstory, traits, relationships). The Locations page lets you design spaces with concept art, color palettes, lighting, props, sound ambiance, and spatial layout.

This metadata pool feeds the AI writing pipeline (logline → concept → treatment → outline → draft) and heuristic systems (shot lists, breakdowns, call sheets, scheduling). The screenplay emerges from the world you build -- or, if you bring an existing script, PRISM catches up by extracting characters, scenes, and locations from it.

5-Stage Hollywood Workflow

Every production moves through a structured pipeline with 20 substages:

AI-assisted evaluation gates check quality before stage transitions, with greenlight leniency allowing one criterion below threshold if the overall score passes.

AI-Powered Creative Tools

PRISM integrates multiple AI capabilities into the production workflow:

• Script Enhancement -- Rewrite and polish dialogue, action lines, and scene descriptions via Gemini
• Script Intelligence -- LLM-based per-scene entity extraction (characters, props, wardrobe, vehicles, SFX/VFX, stunts, extras), mood/complexity scoring, and cross-scene continuity analysis with auto-annotation
• Storyboard Generation -- Scene-to-image prompt engineering produces visual frames via Imagen 4
• Poster Generation -- AI-generated production artwork via Imagen 4
• Draft Generation -- Full screenplay generation from locked metadata (logline, outline, characters)
• Voice Generation -- Text-to-speech table reads via Gemini 2.5 Flash TTS with 29 named voices
• Music Generation -- Score creation via Google Lyria 2 (Vertex AI)
• Video Generation -- Scene-to-video via Hugging Face Inference API (Wan2.2)
• AI Co-Writer -- Story pipeline assistance for concept, treatment, outline, and character development

Configurable AI Pipelines

PRISM supports multiple content types through configurable AI pipelines. Each pipeline defines its own stage sequence, context blocks, and AI prompts. Three seed pipelines are included:

• Film/Hollywood (default) -- Standard screenplay development
• News -- Journalistic content with fact-checking, source tracking, and style guides
• Sales Pitch -- Business presentations with market data, audience profiles, and objection handling

Pipelines are managed through an admin UI with drag-and-drop stage editing and AI-assisted pipeline generation. Eight modular context block providers (world rules, source material, fact check references, news style, market data, audience profile, product data, pitch style) supply domain-specific context to AI generation.

Annotation Overlay System

PRISM implements a "transparent acetate sheet" metaphor for collaborative screenplay annotation. Annotations are layered with four visibility levels:

Seven annotation types are supported: note, question, action item, source, timing, data, and media. Colored gutter indicators in the screenplay margin show annotation density at a glance.

AI Provider Architecture

PRISM abstracts AI providers behind a unified interface with a per-task model routing matrix that maps 8 task categories (enhancement, evaluation, draft, scaffold, intelligence, chat, storyboard, poster) across 6 subscription tiers to specific provider/model combinations.

Text providers: Google Gemini, Anthropic Claude, OpenAI GPT, Cloudflare Workers AI (Llama 4 Scout, Mistral, Gemma), HuggingFace (cost-effective extraction with JSON quality gate and auto-fallback).

Image providers: Google Imagen 4, OpenAI GPT Image, Cloudflare Workers AI (FLUX.1 Schnell).

Specialized providers: Gemini TTS (voice), Lyria 2 via Vertex AI (music), HuggingFace Inference (video).

All AI SDK calls route through Cloudflare AI Gateway for caching, analytics, and rate limiting. Workers AI serves as a zero-cost system fallback requiring no user API key. The BYOK (Bring Your Own Key) cascade resolves: User key → Org key → System token.

Real-Time Collaboration

Socket.IO-based production sync enables live session coordination. The Inspector sidebar provides context-sensitive information based on cursor position -- Scene breakdowns, Character profiles, Story pipeline status, and AI actions. An NLE-style timeline with JKL transport and three-way playhead sync (screenplay hover, timeline scrub, storyboard highlight) keeps all views coordinated.

On-Set Device Integration (Nexus)

The Nexus console is a separate Tauri/Rust application that runs on an ODROID N2+ on set. It communicates with the PRISM server API to provide:

• Live camera capture (V4L2/MJPEG)
• WHIP/RTMPS streaming
• Session and event tracking
• Timecode-synced markers
• Video segment upload with idempotent confirmation
• Screenplay-driven DVR with scene-based navigation
• Multi-cam angle switching (8 camera roles)
• Multi-quality adaptive streaming via Cloudflare Stream (HLS)

Connection hardening includes exponential-backoff retry on all HTTP calls, upload queue limits with back-pressure, capture mode crash recovery, device token keychain storage, and upload confirm retry with crash-recovery files.

Production Management

Productions are organized with 8 built-in department templates (Direction, Camera, Art, Sound, Editorial, AD, Script, Producing). The system supports scene breakdowns with 15 industry-standard category templates, locations with GPS/contacts/logistics, shoot day scheduling with call sheets, and cast-to-actor linking. Each department can have its own annotation layer, and permissions cascade through the organizational hierarchy so that department heads see exactly what they need.

Tech Stack

Content-addressable storage minimizes duplication across script versions, and database migrations are applied automatically during deployment.

Organization Hierarchy

PRISM uses a four-tier hierarchy with cascading permissions:

Tenant (Billing Entity)
  └── Organization (Production Company)
        └── Production (Film / Series)
              └── Script (Screenplay)

Permissions cascade downward: an Organization Admin automatically has access to all Productions and Scripts within that organization. Five role levels control access at each tier -- Owner, Admin, Editor, Viewer, and None. Roles can be explicitly set at any level to override or restrict inherited access.

The Project Wizard provides atomic creation of the full hierarchy (organization + production + script) in a single step, with automatic owner role assignment.

What's Next

Continue with the other getting-started guides:

• Installation -- Set up a local development environment
• Quick Start -- Create your first project and write a screenplay
• Environments -- Local, development, and production server details

For deeper dives, see:

• Architecture Overview -- System design and module relationships
• API Endpoints -- REST API documentation
• Permissions -- Cascade permission system in detail
• Deployment -- Production and development server deployment

See Also

• Dashboard -- Navigating the PRISM interface after setup
• Projects -- Creating and managing organizations, productions, and scripts
• Glossary -- PRISM-specific terminology