Dailies Desk

A DIT in your NAS: turns a shoot day of offloaded footage into a verified dailies report with backup checksums, clip specs, slate transcripts and problem flags. Footage never leaves the building.

Comment

Inspiration

I have been a cinematographer since 2008. The most expensive mistake I have seen on set is not a bad shot. It is a corrupt backup or a wrong frame rate that nobody catches until the edit, when the location, the actors and the light are gone.

Big productions pay a DIT (digital imaging technician) to stop that from happening: checksum every card, log every clip, flag problems before cards are reformatted. Indie crews, documentary teams, wedding videographers and solo shooters usually cannot afford one. Their footage already sits on a NAS. So the NAS should do the DIT pass.

What it does

Dailies Desk is a UGREEN AI NAS Skill. Say "run dailies for today's shoot" and it:

  1. Inventories the shoot folder (video, sound, stills).
  2. Reads codec, resolution, frame rate, duration and timecode for every clip.
  3. Compares the shoot folder against the backup with SHA-256 checksums and flags any MISMATCH or MISSING IN BACKUP right away, while the card can still be saved.
  4. Transcribes spoken slates ("scene 12, take 3, mark") into scene and take labels.
  5. Describes reference stills for continuity.
  6. Writes a dailies report with a ranked problem list: critical, warning, note.
  7. Saves the report where you confirm, and optionally tags clips dailies-verified or dailies-review after you approve.

Unreleased footage, often under NDA, never leaves the building. Everything runs on the NAS and its local models.

How we built it

Dailies Desk follows the official Uliya Skill Development Standard: one SKILL.md plus two linked references (problem flag rules and the report layout). It only delegates to the registered NAS experts (file-agent, knowledge-agent, photo-agent, document-agent), and each step receives only real results from the step before. It handles all five delegation states, never deletes, moves or overwrites footage, and asks before every write. It passes the official skill kit checker with 0 errors and 0 warnings.

Challenges

Writing the workflow so it never guesses. A dailies report that invents a take number or a checksum is worse than no report, so every value must come from a real NAS result, and anything that could not be checked goes into a "what was not checked" section.

What's next

Validate on a real UGREEN AI NAS: real media info fields, transcription of slates inside video files, and checksum speed on full cards.

Built With

Share this project:

Updates

Submission history