Executive summary

What was built

The Cabinet of Minor Gods is a complete sixty-work generative world consisting of:

• 60 canonical 4:5 portraits
• 60 five-second canonical manifestations
• 60 five-second nonverbal audio relics
• 73 final evidence-film speech performances
• 10 procedural chapter scores
• 10 liturgies
• One definitive 11:53 evidence film
• One interactive still/motion/audio Cabinet
• One film-synchronized timeline website

Creative premise

The Cabinet imagines a synthetic folk religion for ordinary systems and failures: unsent messages, temporary construction, unread notifications, forgotten passwords, delayed buses, office copiers, potholes, receipts, cupboards, breakrooms, and dead remotes.

Each god receives a domain, offering, taboo, miracle, portrait, manifestation, sound relic, and evidentiary account. The work combines sincere devotion, deadpan administration, tactile materiality, and unresolved testimony.

TITLES role

TITLES MCP provided:

• Artist-model and generic-model discovery
• Persistent project/session state
• 124 generated image outputs
• 60 successful image-to-video outputs
• 119 speech outputs across five providers/models
• Model/adapter/voice selection
• Execution and output provenance
• Cost-bearing inference

Final TITLES spend: $32.047774.

Local-production role

Local tools provided:

• The sixty-god concept system and manifests
• Human/agent selection and coverage management
• Ten procedural scores
• Sixty procedural audio relics
• Diegetic answering-machine/intercom/witness DSP
• Semantic timing and edit conform
• Typography and website-slot framing
• Ten chapter films and definitive master
• Interactive websites
• Media/browser/deployment verification
• Hosting package and checksums

Major learning

The project did not become complete when generation completed. The decisive artistic improvements happened through review and conform:

• Replacing narration with evidence
• Refusing unnecessary additional visual generation
• Restoring native 4:5 presentation
• Aligning motion to semantic events
• Removing abrupt returns to starting stills
• Transitioning while both works remain in motion
• Translating the website's slot component into film language
• Turning the grid itself into a timeline

TITLES roadmap implication

TITLES is already strong at model discovery, artist-model lineage, and exact generative execution. The largest adoption opportunity is the structured handoff from outputs to authored systems:

• Batch/coverage orchestration
• Project-level spend and provenance ledgers
• Named cast and recording environments
• Editorial timelines and semantic cue placement
• Deterministic sound/effect workflows
• Automated media QA
• Exportable collection and timeline websites
• Bulk sanitized deployment packages

The strategic lesson is not that TITLES must replace FFmpeg, ImageMagick, or web development. It is that TITLES can make the path from generative output to specialist production explicit, portable, and reversible.

Final distribution

The hosting package is a self-contained static artifact with no API keys and no runtime TITLES dependency. It includes the two websites, the complete evidence film, all 181 runtime media files, deployment instructions, provenance, process documentation, and a checksummed manifest.