[OptimistOtaku]

Opening an older captioned project converts every edited-time caption and word timestamp back to source time. Each timestamp currently scans the clip mappings backwards, so heavily edited recordings pay for a clip scan per word. In the committed complete-synchronization benchmark, a 2,000-clip/20,000-word project drops from 130.42 ms to 30.24 ms (4.31× faster).

For at least 128 clips, validate that edited starts and ends are finite and monotonic, then binary-search the last mapping whose start is at or before the timestamp. Shared boundaries retain the previous reverse scan's last-match priority. Small timelines and invalid/non-monotonic mappings retain the original lookup. Rounded final timestamps and fullscreen-hold conversion are unchanged.

This runs through GPUI caption synchronization and the shared editor-loading path. It only accelerates the legacy source_timed = false conversion; already source-timed captions retain their path. No persistent cache, public API, dependencies, or project-format changes.

Benchmark

Windows x86_64, Ryzen 5 5600H, Rust 1.88 GNU, repository release profile (opt-level=s, thin LTO). Seven alternating paired samples after two warmups; five fresh projects per sample. Inputs are cloned outside timing and complete output projects are compared before measurement. Timings include the full synchronize_captions call: migration, caption projection, sorting and output construction. File I/O, model inference, GPUI launch, and output destruction are excluded.

Clips	Words	Before	After	Speedup
1	9,000	8.024 ms	8.293 ms	0.97×
60	1,500	1.447 ms	1.354 ms	1.07×
500	10,000	19.806 ms	9.898 ms	2.00×
2,000	20,000	130.415 ms	30.242 ms	4.31×

The single-clip control was approximately 3% slower. A second complete run measured 1.89× and 4.51× gains at 500 and 2,000 clips, with a roughly 6% slower single-clip control and a neutral 60-clip control. The gain is for older projects with many clip mappings; this is not an overall editor-startup multiplier.

cargo test --locked -p cap-project --release --lib benchmark_caption_migration_lookup -- --ignored --nocapture

Validation

• cargo fmt --all
• cargo check --locked -p cap-project
• cargo test --locked -p cap-project --lib --tests: 204 library tests and four integration tests passed; five benchmarks ignored by default, new benchmark run separately.
• cargo clippy --locked -p cap-project --all-targets -- -D warnings
• git diff --check

The reference synchronization function matches upstream f7d351611 exactly. Tests compare whole migrated projects and repeated synchronization, including styles, reordered clips, transitions, overlapping holds, shared/zero-length boundaries, gaps, neighboring float values, non-finite timestamps and invalid timescales. Existing rounded-timeline-end migration tests pass. Native GPUI startup was not exercised locally.

Confidence Score: 5/5

This PR appears safe to merge; no actionable issues were found.

Summary

Speeds up legacy caption migration by replacing repeated reverse scans with binary searches on timelines with at least 128 ordered clip mappings.

• Large, ordered clip lists map legacy caption times faster.

_{Reviews (1) · Last reviewed commit: "perf(project): retain linear migration f..." · Reviewed by Greptile}

[vercel]

@OptimistOtaku is attempting to deploy a commit to the Cap Software Inc Team on Vercel.

A member of the Team first needs to authorize it.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing.