Go Pug Go

Go Pug Go is an animated pug short-form series built in Moho for YouTube and short-form platforms, with Ernest as showrunner / lead animator and a small bench of freelance riggers and animators. Episodes run roughly 1–5 minutes each. The catalogue is already four series deep — Pug Impressions, Do Pugs, Random, and Ancillary — with 71 Pug Impression concepts written down.

Alongside the series, we're building Ernest's production infrastructure from scratch: a custom web application running on the existing VPS that handles incubation (idea-stage), greenlight, production tracking, video-frame-accurate review, file-handoff, budget, and audience analytics. It's the system the studio runs on, not just a tracker.

The incubation surface is a single sortable table. Columns mirror the existing spreadsheet (Anim Name, Story, Animation Notes, Dialogue Notes, Music Notes, Budget Tier, Status) and extend it with reference links, cast list (other pugs / props / animals appearing), tags (slapstick, wholesome, gross-out, dance, seasonal), and a status richer than greenlit-or-not (ideated → written → sketched → ready-to-greenlight → greenlit → archived).

The status column is the moment Ernest's mental model lives in the system. The greenlight button on a row is the moment the system earns its first real value: one click forks the row into a full production project with scenes, takes, notes, budget, and assignments — while the incubation record stays in place as the historical brief.

The dashboard is the at-a-glance view: cumulative state, what's happening this week, what's blocked on him, what's blocked on a freelancer. Critically, activity status is auto-derived from take state + last-touched timestamps — Ernest doesn't have to update it. Production stage is manually moved; activity is just visible.

This is the surface the system stands or falls on. Ernest scrubs a take, pauses on a frame, clicks where the pug's ear is wrong, types "the ear deflates a frame too early", and the note pins to that (timecode, x, y). The artist opens the take and sees the pins along the timeline; clicks one and the player jumps there with the marker centered on the frame. Threaded replies. Resolvable. Searchable.

Notes have three flavors: timeline-only (pinned to a timecode), spatial (pinned to a position on the frame at a timecode), and take-level (general feedback, no anchor). One data model, three surfaces.

The kanban view is a projection of the production table. Columns are production stages; cards are shorts. Drag a card to move stage. Identical data, different question being asked.

Folder convention per project

One canonical shape, enforced at upload. Every short is a folder with the same skeleton:

go-pug-go/projects/CheetahA/
  moho/                 ← the .moho file lives here
  assets/
    characters/         ← linked, not embedded
    audio/
    reference/
  takes/                ← exported videos for review
  notes/                ← screenshots, PDFs, scribbles
  renders/              ← optional, only on final delivery

Three upload modes

The upload UI offers a clear three-way choice so size + treatment match intent:

Validation at upload

The uploader is a drag-and-drop surface that validates the folder shape before accepting the upload. Does it have a moho/ subfolder? Does it contain at least one .moho file? Does the project name match the system's project ID? If anything's off, the upload is rejected with a specific, actionable error. The validator is the magic. Without it, the convention rots in 3 months — Ernest's tone gets gentler with each freelancer, and standards drift. A machine enforces what a human eventually won't.

Storage math + archive

250GB on the VPS holds roughly 6–9 months of active work before pressure. The archive flow is built in from day one:

• Active project → all takes, all uploads, full versioning kept online
• Completed but not archived → final delivery kept online, intermediate snapshots tarred but kept
• Archived → only the final video + metadata stay online (~50MB). Everything else exported as a single project-CheetahA-full-archive.tar.gz Ernest downloads locally before the system deletes the heavy bytes from the VPS.

The metadata always stays — notes, timeline, takes, status, dates, owners, budget. Just not the binaries. Five years later Ernest can scroll back through every short he ever made; the bytes live on his external drives.

1. Incubation table. Just the table, with filters, search, status field, and Ernest's columns from the spreadsheet. Migrate the existing 71 + 39 + 39 + 37 rows. Make Ernest use it daily.
2. The greenlight transition. Button that promotes an incubation row to a production project. Empty production project structure with scenes-optional, takes, notes, budget fields.
3. The Moho folder validator + uploader. Drag-and-drop with three upload modes (review take / project snapshot / final delivery). Validates the folder shape, zips server-side, stores as a take version.
4. Take review UI. Video player + timeline notes + spatial notes. The only really custom piece; the one the system stands on.
5. Dashboards. Ernest's (review queue + needs-attention) and the producer view (team-load + budget + active projects). Falls out of the data the first 4 steps generate.
6. Archive workflow. Multi-step confirmation, tar export, confirm-then-delete, metadata preserved.
7. Publish log. Per-short release tracking + per-platform performance pull (YouTube, Instagram, TikTok). Becomes the most valuable dataset Ernest owns over time.
8. Freelancer onboarding view. Per-user scoped homepage + assignment view + pinned guidelines.

1–5 is the MVP and ships Ernest a tool that already beats the spreadsheet + email status quo. 6–8 are the durability features. The build phase rolls into the ongoing producer engagement after launch — no separate build invoices, no special-project pricing.