Davinci-ProUser · an open laboratory
We test what Resolve supposedly can't do.
DaVinci Resolve has a scripting API. Everyone repeats the same list of things it "can't" do — and almost nobody has actually checked. We're checking. All of it. Live, in public, with the receipts kept forever.
why this is exciting
If even half of these flip green…
Every directive in our ledger comes from a real shoot, a real edit, a real deadline. These aren't features — they're superpowers waiting on evidence.
Resolve transcribes but won't give the text back. We're hunting the transcript through the project database itself. Find it, and text-driven editing lands inside Resolve — no round-trip, no export.
No API for building node graphs — officially. We inject externally-generated grade files as a smuggling vector. Natural-language grading rides on top.
Ingest, scene bins, proxies, preview LUTs, overnight QC renders — every piece is documented API. The experiment is the chain. Then the chain gets a brain: camera auto-ID.
Resolve is not a live mixer, says everyone. Fusion driven by JSON gets near-live on render. How near? We measure instead of guessing.
A local bridge already connects our web app to a live Resolve. Thumbnail polls, marker punch-ins, seek control — the control surface grows from there.
We dump every object's real methods and diff against the docs. The undocumented surface becomes a public map — our gift to every Resolve scripter.
how the lab works
Science, not vibes.
"Impossible" is a hypothesis
Every claim of "the API can't" goes in the ledger as a falsifiable statement — proven-if / disproven-if, written before any code runs.
Experiments run against a live Resolve
Guarded run.py harnesses in junk projects. Fire them from a button in the Lab app — the bridge carries the call to the real binary.
Evidence or it didn't happen
Stills, API dumps, timing tables, RESULT.md. Failures stay in the record with full honors — a mapped wall is a finding.
The ledger flips in public
🔴 → 🟢 happens where you can watch it. The scoreboard is computed from the data, never hand-tuned.
Findings become tools
Proven surfaces graduate into the app, the in-Resolve panel, and camera-aware recipes. The lab and the product are the same organism.
the scoreboard today
Where the lab stands.
already real
This isn't a pitch deck. It runs.
Mac app + identical web twin. Interactive ledger, experiment kanban, API console, LUT lab — every claim clickable.
A 300-line localhost server (stdlib Python, nothing to install) that lets the app fire real experiments at your running Resolve.
Workflow Integration plugin docked inside Resolve itself — Resolve plugins are Electron apps, so the Lab UI graduates naturally.
Nikon ZR cinema pipeline: 1D normalizations, 3D looks, gamma-aware false color for N-Log and Log3G10.
Nikon ZR + Sony α7III encoded as machine-readable contracts: formats, dual-base ISO, fps matrices, node recipes.
Entire Resolve 20.3.2 + 21 Developer trees, community docs, 66 validated ecosystem resources, zero dead links.
jump onboard
Three ways in.
The ledger and experiment record are live on the web. Every flip, every failure, every timing table — public as it happens.
Download the Lab app, run bridge.py next to your own Resolve, and fire the probes at your setup. Different OS, different Resolve version — your results are data we don't have.
Claim an experiment from the queue. Each one is a self-contained HYPOTHESIS.md + run.py — small enough for an evening, real enough to matter.