Alex CoulombePresents
← Video QA Workbench

Beta field guide · local-first

Better review starts when the render can show its work.

Video QA Workbench is being prepared as an invitation-only beta for teams who need more than a vague quality score—and less than another place their footage can escape.

Today

Local screening, evidence, and strict shot comparison.

Beta promise

A signed local package with a clear renewal path.

Not a claim

No mystery score, automatic approval, or renderer ranking.

How a review works

01

Start with evidence, not a score

A review begins with measurable signals: frozen spans, black-frame candidates, sharpness and motion changes, cue alignment, and context around cuts. The workbench turns those into seekable timestamps and evidence frames. An alert is a reason to inspect, never a claim that a render is objectively good or bad.

02

Keep creative intent visible

Technical signals, team preferences, creative candidates, ambiguous items, and timeline context are intentionally separate. A camera-speed house rule can be useful without pretending it is an objective defect. A human reviewer keeps the final call.

03

Compare the same shot, honestly

For aligned stills and matched revision frames, the workbench maps changed regions, luma, color, and edge structure. The Shot matrix compares each platform only against its matching camera shot. It will not resize, auto-align, or name a renderer winner; wrong framing is evidence worth seeing.

04

Keep the media under your control

The review worker, source playback, reports, and evidence stay local by default. Feedback records dispositions and missed issues locally. Any future telemetry or case-study contribution must be explicit, limited, and separate from the act of reviewing confidential footage.

05

Make feedback improve the next review

Confirm, dismiss, mark intentional, flag a preference disagreement, or log a missed issue. The useful learning signal is not raw upload volume; it is a controlled record of what was helpful, ambiguous, or absent, tested again on separate clips before a behavior changes.

Important limits

A useful beta should say where it stops.

  • It is not an automated quality certification or a substitute for a creative review.
  • It does not infer why two renders differ from pixels alone.
  • It does not turn platform comparison into a single winner score.
  • It does not silently repair camera, crop, pose, or timing mismatches.

Availability

The signed beta is not open yet.

The first invitation wave opens only after the package, renewal path, clean-machine install, and support documentation are ready. If you review real renders repeatedly, you can raise your hand now.

Join the early-access list →Read the tester handbook →