[miguel-heygen]

What

normalize-audio --lufs no longer cuts a clip that is quieter than the target. When the true-peak ceiling stops a boost, the clip keeps its current gain and limitedBy says why.

Why

A clip measuring -17.1 LUFS with a peak past the -1.5 dBTP ceiling, asked for -17, was written with data-volume 0.81. That projects to -18.9 LUFS: further from the target than before, reported as ok: true. The limit math took the lowest of the wanted gain and the ceilings, and the peak room is negative for such a clip, so a boost request became a cut.

Related work

Trim windows are measured correctly on main (-ss/-t as input options); this is not a window bug. Earlier work: the playback-start fix and the in-scene targeting fix, both merged.

How

absoluteLoudnessPlan clamps the limited gain so that a request to go up never lands below the clip's current gain. Cuts are unchanged. A held clip has changeDb 0 and is not reported as matched: absoluteOutcome says matched, held or change, the command prints that it cannot get closer, and nothing is written. This also explains a clip reported withinTolerance: true while far from the target.

Test plan

• Unit tests added/updated
• Manual testing performed
• Documentation updated (if applicable)
• Comments follow CONTRIBUTING.md "Comments"

fails without the fix: with the clamp removed, never cuts a clip that is below the target because its peak is already past the ceiling fails with expected -1 to be close to +0.

What I measured

loudness.test.ts: 47 passed on three runs on the devbox; the cli normalize tests pass (33). Without the fix, 1 failed; with absoluteOutcome forced to matched, calls a clip a limit cannot move held, not matched fails with expected 'matched' to be 'held'. Real ffmpeg on a built CLI: a pulsed tone at -17.1 LUFS with a near full-scale peak, --lufs=-17 --write --json, wrote 0.81 and projected -18.9 LUFS before the fix. A trimmed clip with the same tone measures its played window correctly (-21.9 LUFS).

What I did NOT execute

The fixed CLI on the pulsed fixture after the change, and the reported "-19.7 LUFS, within tolerance, wrote nothing" case, which I could not tie to a cause (an earlier gain already on the clip would produce it).

Size

Under 100 lines on purpose: normalize-audio cuts clips it should boost today, and no open change touches it to carry this.

[heygengenesis]

Verdict: REQUEST CHANGES · 1 blocker · 1 important · 0 nits · head 431a9e9c

The fix itself is right. The fails without the fix: line holds: with the clamp removed, the new test fails (expected -1 to be close to +0), and 44/44 pass at the head. cli normalize-audio tests pass 33/33.

• blocker: Comments is red. The run reports RAISED comment share: packages/studio-server/src/helpers/loudness.ts is 5.0% (28/563), at the base 4.8% (27/560). That comes from the one new comment at line 547. Cut it, or move the explanation into a name (for example const boostOnlyGainDb = …). ↓ inline
• important ↓ inline: a held clip reports withinTolerance: true and prints "Already matched" however far it is from the target.

Scope

Read all of absoluteLoudnessPlan and both callers (CLI normalize-audio.ts and the Studio /loudness/normalize route). CI and Windows were still running when this was posted.

[github-actions]

Edit accuracy: accurate 271 (base branch 271), smooth 141 of those

The gate passes.
Smoothness is reported in the artifact, not gated. A case fails only if it fails 2 of 3 runs.

Quarantined, measured but not gated (0)

Unstable (3)

• seqplainkeys-none-center-r0-nested-z100: tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo false, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true
• seqresize-none-px-r0-nested-z100: tracking 0.06, pressJump 0, drop 0, reload 0, render 0.02, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.02, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.02, renderKey -, undo true, teleport true
• seqplainkeys-none-pct-r0-root-z100: tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo false, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true / tracking 0.06, pressJump 0, drop 0, reload 0, render 0.03, renderKey -, undo true, teleport true

[heygengenesis]

Verdict: REQUEST CHANGES — 1 resolved, 1 open, 2 new · 431a9e9 → e3b0605

New blockers, both from CI on this head: Comments is red again (loudness.ts 28/572, base 27/560), and oxfmt --check fails on normalize-audio.ts, which turns the required Format and Windows preflight checks red. Running oxfmt on that file fixes the second. Locally: loudness 47/47, cli normalize-audio 33/33.

[heygengenesis]

Verdict: APPROVE — 4 resolved, 0 open, 0 new · e3b0605 → fbea11d. Loudness tests pass 47/47 and cli normalize-audio 33/33 locally; oxfmt is clean; CI, Windows, Comments and CodeQL are green.