First Pass checks a master against a destination spec before you submit. Values are evaluated from the master's technical metadata; ffmpeg's ebur128 filter measures those values from real media; the demo masters ship as authored metadata.
pure Python · stdlib only · the model computes nothing here
how far each language sits from this destination's loudness target
what to fix, and in what order — longest lead time first
| CLAUSE | SEVERITY | MEASURED | EXPECTED | LANGUAGE | FINDING |
|---|
how close each language version is to shippable
| LANGUAGE | READINESS | SCORE |
|---|---|---|
| No language scores yet. Click RUN to check. | ||
one bounded agent · four allowlisted MCP tools
On a PASS the agent writes nothing.
| TIME (UTC) | OPERATION | DETAIL | LINK |
|---|
Industry commentary from post-production delivery specialists puts first-submission QC failure in the 20–30% range — a directional figure, not an official published statistic. Most often the cause is mundane and preventable: an audio mix delivered at theatrical loudness (~−24 LUFS) against a streaming platform spec requiring ~−27 LUFS. Every rejection incurs redelivery fees and risks missing an announced premiere date.
Master metadata JSON
Destination spec JSON
(StreamOne or HallArc — both fictional)
Evaluates LUFS, True Peak, HDR, subtitles, packaging, CBFC gating from the master's technical metadata. LLM never computes.
Four MCP writes only:
Verdict banner · ranked fix plan · per-language readiness · MCP action ledger
One bounded agent, four allowlisted tools, deterministic Python that the model never touches.
The finished film file you are about to deliver. It carries technical metadata about the picture, every audio language, the subtitles and the packaging.
The delivery rules a platform publishes: how loud the mix must be, which colour standard, which subtitles are required. StreamOne and HallArc are two FICTIONAL destinations authored for this project, with deliberately different targets — which is why the same master can pass one and fail the other.
LUFS measures perceived loudness over a whole programme. A cinema mix usually sits near -24 LUFS; streaming platforms typically want around -27, so a theatrical mix delivered unchanged fails. The scale here is RELATIVE: 0 LU is this destination's target, so a bar left of centre is quieter than required and right of centre is louder. dBTP is the loudest instantaneous peak — exceed the ceiling and the audio can distort on playback.
A blocker means the platform will reject the file; a warning will not stop delivery. Jobs are ordered by how long the fix takes to arrange, not by pipeline order — the slowest thing to start is listed first.
How close each language version is to shippable. For a simultaneous multi-language release, certification of the ORIGINAL language gates clearance of the dubs, so one pending certificate can hold up every other version.