[chiragjangiir]

Description

src/lib/ffmpeg.ts held a second, complete copy of the FFmpeg argv builder (buildVideoFilter, buildAudioFilter, buildArguments) that nothing imported — buildArguments is not exported and has zero references in the repository at main. The unit tests imported that copy, so they stayed green while src/lib/ffmpeg.worker.ts ran its own copy that had silently drifted behind it.

Drift between the tested copy and the executed copy:

Behaviour	ffmpeg.ts (tested, never called)	ffmpeg.worker.ts (executed)
Sharpness setting	unsharp=5:5:n:5:5:0.0	missing — the slider had no effect on exports
Volume setting	volume=n.nn	missing — the slider had no effect on exports
trim with null trimEnd	trim=start=X (FFmpeg infers the end)	trim=start=X:end=999999 placeholder
atrim with null trimEnd	atrim=start=X	atrim=start=X:end=999999 placeholder
-t output cap	unconditional (PR #1289)	gated on recipe.speed !== 1

The last row is the user-visible one: PR #1289 ("fix: enforce -t flag on export to prevent infinite loop with background music") is merged, but its diff touched only src/lib/ffmpeg.ts — the uncalled copy. The executed path kept its speed !== 1 gate, so the fix never ran at runtime, issue #1577 was closed without a behavioural change, and the duplicate report #1493 is still open.

What changed

• New src/lib/exportGraph.ts: one pure implementation of the export graph — buildVideoFilter, buildAudioFilter, buildAudioTrimFilter, buildDurationArgs, buildArguments. No worker globals, no DOM, no FFmpeg instance.
• src/lib/ffmpeg.worker.ts imports it and deletes its private copy; src/lib/ffmpeg.ts deletes its dead copy and re-exports the two filters the existing tests import.
• buildDurationArgs centralises the -t decision and rejects durations that cannot be trusted (0 from a failed metadata load, Infinity from MediaRecorder WebM containers, NaN), which would otherwise reach FFmpeg as -t 0 (zero-length export) or -t Infinity (aborts the export).
• New src/lib/tests/exportGraph.test.ts (22 tests) asserts the argv the worker actually passes: duration capping, the looped-music case, volume and muting, codec-per-container, overlay positioning, and the removal of the 999999 placeholders.

Net effect: the export behaves the way the reviewed code intended — Sharpness, Volume and the -t cap are live — and the tests can no longer pass against a copy nothing runs.

Related Issue

Closes #1493

Related: #1577 (closed by #1289 without a behavioural fix), #1289 (merged, landed in the uncalled copy), #1590 (closed unmerged).

Type of Contribution

• Bug fix
• New feature
• Documentation update
• Refactor
• GSSoC contribution

Participant Info

• GitHub username: chiragjangiir
• Contribution level (Beginner/Intermediate/Advanced): Advanced

Screen Recording

A recording is being prepared and will be attached to this PR. Suggested demonstration: bun run dev → Audio & Speed → Volume = 0% → Export → the exported file's audio is silent (on main the same export keeps full-volume audio).

Checklist

• I have read the contribution guidelines
• My changes follow the project structure
• bun run lint passes — run as npx eslint . → exit 0 (bun is not installed on this machine)
• bunx tsc --noEmit passes — run as npx tsc --noEmit → exit 0
• bun run build passes — run as npx next build --webpack → exit 0
• New interactive elements have aria-label / accessible names (no new UI added)
• No console.log statements left in
• This PR is related to a valid issue ([BUG] Export Can Hang Indefinitely When Video Has No Audio and Background Music Is Looped #1493 — open, unassigned, no linked PR)
• Tested in Chrome, Firefox and Safari (Chrome verified end to end; Firefox/Safari not run)
• Screen recording attached above (required for UI/feature changes)

How this was verified

Unit tests

$ npx vitest run src/lib/tests/exportGraph.test.ts src/lib/tests/ffmpeg.test.ts src/lib/tests/videoFilter.test.ts
 ✓ src/lib/tests/exportGraph.test.ts (22 tests)
 ✓ src/lib/tests/ffmpeg.test.ts (9 tests)
 ✓ src/lib/tests/videoFilter.test.ts (16 tests)
 Test Files  3 passed (3)      Tests  47 passed (47)

Static checks: npx tsc --noEmit → 0 · npx eslint . → 0 · npx next build --webpack → 0 (static export generated).

Runtime A/B through the real editor UI. Two dev servers: localhost:3100 = this branch, localhost:3101 = main. Each run: import a ~3 s fixture → set Volume → EXPORT → decode the exported MP4 in the page and measure the audio.

Build	Volume	Source RMS	Exported RMS	Ratio	Export time
main	0%	0.35235	0.35085	0.996 (setting inert)	~60 s, progress readout showed -287106300%
branch	50%	0.35080	0.17535	0.500	2 s
branch	0%	0.35080	0.000000 (peak 0)	0.000	2 s

On main the Volume slider has no effect on the exported audio; on this branch it is exact — 50% halves the amplitude and 0% produces digital silence.

The -t cap was confirmed in the live argv with a temporary diagnostic (since removed): "-t", "3.004000" for a 3.004 s fixture produced a 2.963 s file — no truncation, no overshoot.

Pre-existing test failures (not introduced here). npx vitest run reports 4 failing files / 6 failing tests out of 131; all four fail identically on an unmodified main worktree: src/lib/exportEstimate.test.ts (duplicated expect import — parse error), useVideoEditor.test.ts and ThemeToggle.test.tsx (react 19.3.0 vs react-dom 19.2.8 mismatch from f013f3c), and editorPersistence.test.ts (localStorage undefined). CI does not run vitest (main.yml runs install + build only).

Limitations

• The looped-background-music scenario from [BUG] Export Can Hang Indefinitely When Video Has No Audio and Background Music Is Looped #1493 was not reproduced end to end (no music file could be attached in the verification environment). What is verified is that -t now reaches the executed argv whenever the duration is trustworthy — exactly what fix: enforce -t flag on export to prevent infinite loop with backgrou… #1289 intended, and what the new tests assert.
• The ~60 s vs 2 s export time above compares an MP4 fixture (branch) against a WebM fixture (control); treat the timing gap as suggestive rather than causal. The duration/-t evidence is container-independent.
• Structures are unchanged: no new dependencies, no UI changes.

[netlify]

👷 Deploy request for reframe-os pending review.

Visit the deploys page to approve it

Name	Link
🔨 Latest commit	54e0b7a

[github-actions]

👋 Thanks for your PR, @chiragjangiir!

Welcome to Reframe — a browser-based video editor built for everyone 🎬

🟠 GSSoC'26 PR detected — thanks for contributing under GirlScript Summer of Code 2026!

What happens next

1. 🤖 Automated checks — build & TypeScript typecheck will run automatically
2. ✅ Vercel preview — a preview deployment will be created (requires maintainer authorization for fork PRs)
3. 👀 Code review — a maintainer will review your changes
4. 🚀 Merge — once approved, your PR will be merged!

Quick checklist

• PR title follows Conventional Commits (e.g. feat: add dark mode)
• Linked the issue this PR closes (e.g. Closes #123)
• Tested the changes locally (bun run dev)
• Build passes (bun run build)

Useful links

• 📖 CONTRIBUTING.md
• 🐛 Open Issues
• ⭐ Star the repo to support the project!

Happy coding! 🎉

[github-actions]

✅ PR Format Check Passed — @chiragjangiir

Basic format checks passed. A maintainer will review your code changes.

This does not mean the PR is approved — it just means the format is correct.