# Merge #2: my five pre-upload tools are now one suite

> Source: <https://dev.to/singhularityai/merge-2-my-five-pre-upload-tools-are-now-one-suite-1i5a>
> Published: 2026-11-21 14:02:08+00:00

**Key takeaway:** The suite returns HOLD (exit 2) if any module HOLDs or FAILs and PASS WITH WARNINGS (exit 0) if any warns, and a module with nothing to run on reports SKIPPED rather than counting as a pass.

Last month I audited my own fleet — 44 published open-source tools — and the audit ordered 18 of the quiet ones merged into 4 flagships. Merge #1 turned five AI-agent-setup auditors into preflight-agent. Merge #2, shipping now, is the creator side: **faceless-qc-lite v2.0.0**.

faceless-qc-lite started in October as a free pre-upload checker for faceless videos: loudness, dead air, black frames, caption speed, resolution — five checks on the video file, one PASS/HOLD receipt. It did that job well. It also sat next to five other tools I'd built for the *rest* of the same upload: caption-toolkit, thumbnail-checker, retention-checker, shorts-packager, link-rot. Six installs to answer one question: *is this upload ready?* Now it's one:

pipx install faceless-qc-lite

faceless-qc-lite suite my-video.mp4 --captions my-video.srt --thumbnail thumb.png --title "My Title" --hashtags "#shorts" --catalogue catalogue.json

One run, one receipt. A module with nothing to run on reports SKIPPED, with its reason — a skip is never counted as a pass. Suite verdict: HOLD if anything HOLDs or FAILs (exit 2), PASS WITH WARNINGS if anything warns (exit 0), else PASS.

The merge changed no engine: every engine is byte-identical to its source, verdict vocabularies stay per-module, 21 new tests on top of 220 carried over — 241 total, all green.

The fixture is the demo: examples/make_fixture.py builds a set with one planted problem per module: a very quiet 640x360 video with a 3.5s black gap, a caption cue racing at 138 CPS, a 320x180 thumbnail, a 103-character title, a catalogue with a goo.gl link and an example.com placeholder. The README reproduces the real suite receipt verbatim — HOLD, all six sections, exit 2.

**Watch it run (17 seconds):** the short demo video is a real pre-upload run of the original checker — an upload-ready faceless video measured at −21.6 LUFS against the −14 target and held. That catch is why this tool exists. [Watch the demo](https://github.com/jasveerai88/faceless-qc-lite/blob/main/docs/faceless-qc-lite-demo.mp4)

Honest limits: ffmpeg is a hard requirement for the media measurements. The link audit is structural — it has never loaded your link. A PASS means "none of the checked problems were found" — not "this video will perform". It's MIT, zero dependencies (stdlib only), local only, no telemetry, and the anchor stays free forever.

Repo: [https://github.com/jasveerai88/faceless-qc-lite](https://github.com/jasveerai88/faceless-qc-lite)

Feedback I'd most value: uploads the suite misjudges — a failure mode none of the six checks sees, or a verdict the measurements don't support.
