Gluoon Archive

01 — The problem

Media Offline the two words that end a re-grade request

Resolve is the best post tool ever built. It has no memory.

A project from three years ago opens. Half the timeline is red. The footage is somewhere — an LTO tape, a shelf drive, a NAS that got reorganised, a folder called FINAL_v3_new. Nothing in Resolve can tell you where it went. Nothing in Resolve can tell you what was even in the shot.

Power Bins help inside one database, for one user. They are pointers to files: no state, no versions, no idea whether the file still exists. And every keyword, rating and note you added lives inside the project — close it, and the knowledge goes with it.

"There is no way to search for clips or timelines across multiple projects. You can only search one project at a time."
Blackmagic Design user forum

So every facility solves it the same way: a spreadsheet, and one person who remembers. That person is the media asset manager, and they are a single point of failure.

02 — The product

A bridge, not another silo.

Every asset manager on the market asks you to leave Resolve, import your media into their world, and manage it there. Gluoon Archive speaks Resolve's own database, natively, in both directions.

Power Bins, in a browser

Your real bins. Live. No import step.

The app attaches to the Resolve project server's PostgreSQL and reads the pool, the folder tree, the bin colours and the clip names as they are right now — not a mirrored copy that drifts out of date.

Create, rename, recolour, move and delete bins — written back into Resolve
Every write is snapshotted to an audit table before it runs, so it is reversible
Offline clips flagged with a MISSING badge, per bin and per clip
Hover a clip to scrub it, without opening Resolve

archive — Power Bins

Power Bins browser: pool, folder tree with colour-coded bins, and a grid of clips.

The viewer

Frame-accurate, even when the original is offline.

Every archived clip carries a 50-frame scrub strip, each frame stamped with its real timecode. Drag across the image and you are moving through the shot — no media, no mount, no Resolve.

Mark in and out with I and O, exactly as you would in an NLE
Generate a CMX 3600 EDL from your selection, or download the proxy
Filter by tag — brand, channel, or anything you define
Filter to only missing and see what the archive has lost

archive — Viewer

Clip viewer: a large scrubbable frame with timecode, in/out markers, tag filters and source metadata.

GARCH — the transcode farm

The workstations are idle at night. We made them the farm.

Thirteen editing machines — Windows and macOS — double as GPU transcode nodes. They cut 6K BRAW into H.265 proxies and scrub frames, two jobs each, dispatched to whoever is least busy.

Live GPU load, VRAM and temperature per node, in the browser
Pause a node the moment an artist sits down at it
A node dies mid-encode and the task simply re-enters the queue
Retry just the thumbnails without re-encoding the whole clip

archive — GARCH

GARCH dashboard: per-node GPU and VRAM bars, job queue and conversion stages.

03 — The round trip

Find it in a browser. Land it on the colorist's timeline.

A producer scrubs an archived clip in a browser tab, marks in and out, and hits Insert. The subclip appears in the Media Pool of the Resolve instance that is open right now, on the colorist's machine, at frame-accurate timecode.

HOW

A small bridge agent

Runs on the artist's workstation, auto-discovered over the LAN and authenticated with a shared secret. It reports whether Resolve is running, which version and which project is open — so the browser shows a live status dot before you ever click.

WHAT SHIPS

Original or proxy, your call

Send the 6K BRAW if the drive is mounted, or the H.265 proxy if it isn't — same in and out points either way. Drop it into a bin, or straight onto the open timeline. Or skip Resolve entirely and export an EDL.

04 — Architecture

Three layers. All of it on your own hardware.

Layer 01 · Index

Resolve project server

PostgreSQL, attached read and write
Live Power Bin tree — no mirroring
Audit row written before every mutation
Media-offline detection across the whole library

Layer 02 · Compute

GARCH farm

13 GPU nodes, Windows and macOS
BRAW → HEVC on hevc_nvenc
Redis queue, least-loaded dispatch
Optional LUT baked into the proxy

Layer 03 · Access

The browser

Proxies and frames on object storage
Three-tier cache: disk → Redis → S3
Scrub, tag, trim, mark in and out
Export EDL, or push back into Resolve

NAS / LTO → GARCH node → H.265 + 50 frames → object storage → browser → trimmed subclip → live Resolve timeline

Resolve doesn't keep the file path in a column. It keeps it in a blob.

This is why no third party has done this properly. Media paths, framerate, resolution and the user-visible clip name live inside opaque bytea blobs — a serialised Qt stream. You cannot UPDATE a path with SQL. You have to understand the format.

// BtVideoInfo.Clip — decoded, edited, re-encoded, byte-exact
// round-trip verified on 2,000 / 2,000 production rows

  version    1            QDataStream, big-endian
  Path       "Y:\2022\SPOT\A001_0612.braw"            ← dead
  Path       "/Volumes/main/2022/SPOT/A001_0612.braw"  ← repaired
  FrameRate  23.976       a little-endian double inside a big-endian blob
  Resolution 6144 × 3456
  ClipFormat 8203         FOURCC 'BRAW'

We wrote a full decoder and encoder for that format. That is what makes bulk path repair — pointing thousands of dead clips at their new home in a single pass — actually possible.

05 — Not a prototype

Running in production, on real footage, today.

2,993

clips indexed

Every one bound to a live Power Bin.

1,662

BRAW clips

Already the majority of what we archive.

42.2 h

of footage

4,551,401 frames, each addressable by timecode.

541 GB

of H.265 proxies

Browsable while every original sits offline.

And the library it points at — the production Resolve project server this was built against — is why the problem is worth solving at all:

321,381

clips in the Resolve DB

Across every project, going back years.

42,079

bins

Most have not been opened since delivery.

1,918

projects

One search box away from being assets again.

06 — A note for Blackmagic Design

You already built ninety percent of a MAM.

Cloud storage. A web media browser. Review and approve. And now AI search across objects, faces and dialogue. Every piece is there — and every piece stops at the edge of the open project.

"Search In: All Bins: Searches all bins in the project."
DaVinci Resolve 21 New Features Guide — the IntelliSearch toolbar

The AI index lives inside the project. Start a new one and the same files must be analysed again. Meanwhile Archiware P5 — the LTO backbone of post — lists integrations with eleven asset managers. DaVinci Resolve is not among them, because there is nothing on the Resolve side to integrate with.

Three of Resolve's seven pages are companies you acquired and made free: Color, Fusion, Fairlight. This is the fourth.

The question an archive has to answer	Power Bins	IntelliSearch	Gluoon Archive
Search every clip across all projects	no	project only	yes
Look at footage without opening Resolve	no	no	browser
Look at footage when the original is offline	no	no	yes
Keep metadata after the project closes	no	no	yes
Repair every broken media path in bulk	manual	no	one pass
Give a producer access without a Resolve seat	no	no	a URL

DaVinci remembers the project. Nobody remembers the footage.