Open source CapCut alternative.
TypeScript
test(site): the preview perf gate reads the low-end profile
David Lec4dc512c893d
diff --git a/docs/guides/cut/README.md b/docs/guides/cut/README.mdindex 0fb386e5..88550633 100644--- a/docs/guides/cut/README.md+++ b/docs/guides/cut/README.md@@ -52,6 +52,8 @@ The playhead lives apart from the document. It moves sixty times a second and thWhether any of this is true of a given build is measured rather than assumed: `npm run eval:cut-perf` drives a real editor in a real browser against fixture montages and reports how long the picture takes to answer a moved playhead, whether any frame crossing a cut was the wrong one, and whether a fast-cut montage replayed for minutes still holds the picture to the clock the way its first pass did. Its committed report is the baseline a change is diffed against.+The machine it measures on is part of the measurement. A developer desktop has decode slots to spare and never falls behind, so the preview's caps never bind there and a change that would ruin a laptop reads as a clean sheet. `--machine` puts the page on a smaller one: it takes away hardware decode slots, so decoders past the limit deliver their frames late the way software decoding does, and it slows the main thread to a fraction of this one. `npm run eval:cut-perf-lowend` is the profile the reports come from. What the gates ask there is what a viewer would ask — did the frames arrive, did the picture hitch a quarter second behind the sound, and was the last pass as good as the first — and each machine has its own budget for those, because a number that is honest on a desktop is fantasy on a fifteen-watt laptop.+## Color gradingA clip's color is one field on the clip: an optional preset underneath and manual adjustments on top — sliders, tone curves, color wheels, per-hue bands. Every part is absent when neutral, so an untouched clip carries nothing, and the preset's intensity scales its parameters toward neutral so half-strength means half the move, never a crossfade through gray.diff --git a/evals/cut-perf.latest-report.json b/evals/cut-perf.latest-report.jsonindex 25e38bff..ce920b02 100644--- a/evals/cut-perf.latest-report.json+++ b/evals/cut-perf.latest-report.json@@ -1,5 +1,15 @@{"schema": "cut-perf-eval/v1",+ "machine": {+ "name": "desktop",+ "cpu": 1,+ "slots": 0,+ "softwareMs": 0,+ "lagP95S": 0.05,+ "stallShare": 0.002,+ "lateShare": 0.02,+ "decayFloor": 0.02+ },"gate": {"scrubP50Ms": 10,"scrubP95Ms": 16.7,@@ -11,7 +21,10 @@"lagP95S": 0.05,"lagMaxS": 0.25,"decayRatio": 1.5,- "warmMb": 400+ "warmMb": 400,+ "stallShare": 0.002,+ "lateShare": 0.02,+ "decayFloor": 0.02},"fixture": {"clips": 6,@@ -24,10 +37,10 @@"pass": true,"notes": [],"scrub": {- "p50": 7.5,- "p95": 15.8,- "mean": 8.59,- "max": 21.5,+ "p50": 7.4,+ "p95": 15.6,+ "mean": 8.45,+ "max": 24.1,"n": 41},"longTasks": {@@ -37,7 +50,8 @@"max": 0,"n": 0},- "liveSamples": 60+ "liveSamples": 81,+ "machine": "desktop"},{"name": "drag-across-a-cut",@@ -45,10 +59,10 @@"pass": true,"notes": [],"scrub": {- "p50": 7.5,- "p95": 16.1,- "mean": 8.8,- "max": 19.6,+ "p50": 7.4,+ "p95": 15.7,+ "mean": 8.2,+ "max": 16.4,"n": 41},"longTasks": {@@ -58,7 +72,8 @@"max": 0,"n": 0},- "liveSamples": 52+ "liveSamples": 77,+ "machine": "desktop"},{"name": "drag-across-a-transition",@@ -66,10 +81,10 @@"pass": true,"notes": [],"scrub": {- "p50": 7.5,- "p95": 15.8,- "mean": 8.59,- "max": 22.9,+ "p50": 7.4,+ "p95": 15.7,+ "mean": 8.34,+ "max": 17.9,"n": 41},"longTasks": {@@ -79,7 +94,8 @@"max": 0,"n": 0},- "liveSamples": 63+ "liveSamples": 64,+ "machine": "desktop"},{"name": "jump-across-the-cut",@@ -87,10 +103,10 @@"pass": true,"notes": [],"scrub": {- "p50": 15.9,- "p95": 24,- "mean": 15.6,- "max": 25.6,+ "p50": 15.5,+ "p95": 18.2,+ "mean": 14.96,+ "max": 23.7,"n": 41},"longTasks": {@@ -100,7 +116,8 @@"max": 0,"n": 0},- "liveSamples": 88+ "liveSamples": 71,+ "machine": "desktop"},{"name": "play-hard-cuts",@@ -108,7 +125,7 @@"pass": true,"notes": [],"drops": 0,- "presented": 2076,+ "presented": 2075,"boundaryDrops": 0,"longTasks": {"p50": 0,@@ -117,7 +134,8 @@"max": 0,"n": 0},- "liveSamples": 60+ "liveSamples": 60,+ "machine": "desktop"},{"name": "play-with-transitions",@@ -125,7 +143,7 @@"pass": true,"notes": [],"drops": 0,- "presented": 2076,+ "presented": 2074,"boundaryDrops": 0,"longTasks": {"p50": 0,@@ -134,7 +152,8 @@"max": 0,"n": 0},- "liveSamples": 60+ "liveSamples": 60,+ "machine": "desktop"},{"name": "play-fast-cuts",@@ -142,14 +161,15 @@"pass": true,"notes": [],"drops": 0,- "presented": 1619,+ "presented": 1617,"lag": {"p50": 0.02,"p95": 0.03,"mean": 0.02,"max": 0.03,- "n": 1619+ "n": 1617},+ "stallShare": 0,"decay": {"first": 0,"last": 0@@ -164,7 +184,8 @@"liveSamples": 120,"liveSources": 12,"keptSources": 20,- "warmMb": 354+ "warmMb": 354,+ "machine": "desktop"},{"name": "play-fast-cuts-long",@@ -180,6 +201,7 @@"max": 0.03,"n": 6481},+ "stallShare": 0,"decay": {"first": 0,"last": 0@@ -194,7 +216,8 @@"liveSamples": 120,"liveSources": 12,"keptSources": 20,- "warmMb": 354+ "warmMb": 354,+ "machine": "desktop"},{"name": "idle-costs-nothing",diff --git a/site/package.json b/site/package.jsonindex 2dd6abd7..3a9a7477 100644--- a/site/package.json+++ b/site/package.json@@ -28,6 +28,7 @@"test:genvideo": "bun run scripts/genvideo-selftest.ts","eval:genvideo": "bun run scripts/genvideo-eval.ts","eval:cut-perf": "bun run scripts/eval-cut-perf.ts",+ "eval:cut-perf-lowend": "bun run scripts/eval-cut-perf.ts --bucket playback --runs 3 --machine ryzen-5500u","eval:cut-sync": "bun run scripts/eval-cut-sync.ts","eval:cut-export": "bun run scripts/eval-cut-export.ts"},diff --git a/site/scripts/eval-cut-perf.ts b/site/scripts/eval-cut-perf.tsindex 9356c707..c8197b09 100644--- a/site/scripts/eval-cut-perf.ts+++ b/site/scripts/eval-cut-perf.ts@@ -98,6 +98,10 @@ interface Machine {* often it happens says whether the picture hitches; the worst single frame* says only that a seek happened. */stallShare: number;+ /** Late-frame share below which a rise across the play is noise. A machine+ * already dropping a few frames swings by a few frames run to run, and a+ * ratio between two small numbers reads that as decay every other run. */+ decayFloor: number;/** Share of played frames that may show a picture which does not belong at* that instant. This is the plainest reading of "does the video play": a* frame the decoder failed to deliver on time is a frame the viewer sees@@ -108,15 +112,15 @@ interface Machine {const MACHINES: Machine[] = [// The machine this is being written on: nothing binds, and the numbers are// held to the tightest budget in the file.- { name: "desktop", cpu: 1, slots: 0, softwareMs: 0, lagP95S: 0.05, stallShare: 0.002, lateShare: 0.02 },+ { name: "desktop", cpu: 1, slots: 0, softwareMs: 0, lagP95S: 0.05, stallShare: 0.002, lateShare: 0.02, decayFloor: 0.02 },// A mainstream laptop with an integrated GPU.- { name: "laptop", cpu: 4, slots: 6, softwareMs: 6, lagP95S: 0.06, stallShare: 0.02, lateShare: 0.06 },+ { name: "laptop", cpu: 4, slots: 6, softwareMs: 6, lagP95S: 0.06, stallShare: 0.04, lateShare: 0.1, decayFloor: 0.1 },// The one people report from: a 15W six-core with Vega graphics, running a// browser that is also drawing the editor.- { name: "ryzen-5500u", cpu: 10, slots: 4, softwareMs: 8, lagP95S: 0.06, stallShare: 0.03, lateShare: 0.06 },+ { name: "ryzen-5500u", cpu: 10, slots: 4, softwareMs: 8, lagP95S: 0.06, stallShare: 0.03, lateShare: 0.05, decayFloor: 0.08 },// Worse than anything anyone has reported, which is the point: what holds// here holds on the machines that have not written in yet.- { name: "potato", cpu: 20, slots: 2, softwareMs: 12, lagP95S: 0.3, stallShare: 0.12, lateShare: 0.32 },+ { name: "potato", cpu: 20, slots: 2, softwareMs: 12, lagP95S: 0.1, stallShare: 0.06, lateShare: 0.2, decayFloor: 0.15 },];const MACHINE_NAME = arg("--machine") ?? "desktop";@@ -163,9 +167,9 @@ const BASE_GATE = {// Sound is scheduled against the clock the picture is drawn from, so how far// the picture sits behind the source moment it was asked for is how far it// sits behind the sound. A frame either side of the ask is the picture- // keeping up; a quarter second is a viewer noticing.+ // keeping up; a quarter second is a viewer noticing. Each machine profile+ // sets its own, since what is honest on a desktop is fantasy on a laptop.lagP95S: 0.05,- lagMaxS: 0.25,// A long play must end the way it began: a session that decays is the bug// this case exists to catch.decayRatio: 1.5,@@ -182,6 +186,13 @@ const GATE = {lagP95S: MACHINE.lagP95S,stallShare: MACHINE.stallShare,lateShare: MACHINE.lateShare,+ decayFloor: MACHINE.decayFloor,+ // A long task is a budget on work, and the throttle divides how much work+ // fits in a millisecond. Thirteen milliseconds of compositing is thirteen+ // milliseconds of compositing whether the CPU runs it in sixteen or in a+ // hundred and sixty, so the budget scales with the machine and keeps meaning+ // the same thing.+ longTaskMs: BASE_GATE.longTaskMs * MACHINE.cpu,};/** How far behind the sound the picture has to be for that frame to read as a@@ -359,9 +370,18 @@ const SHOT_INS = Array.from({ length: SHOTS }, (_, i) => {});/** Seed the montage: one asset, many short clips, each its own mapping. */-const seedMontage = async (page: Page): Promise<Fixture> =>+/**+ * The montage, optionally wearing what a real project wears.+ *+ * The bare fixture is footage and cuts, which measures the decoders and very+ * little else. A cut people actually make carries a look and a grade, and both+ * are per-frame work on the compositor's side of the fence — the side a weak+ * CPU runs out of first. Measuring only the bare one leaves the cost that+ * shows up in the reports unmeasured.+ */+const seedMontage = (dressed = false) => async (page: Page): Promise<Fixture> =>page.evaluate(- async ({ ins, shot, srcDuration }) => {+ async ({ ins, shot, srcDuration, dressed }) => {const dev = (window as unknown as {__cutDev: { useEditor: { getState(): Record<string, unknown>; setState(p: unknown): void } };}).__cutDev;@@ -385,6 +405,13 @@ const seedMontage = async (page: Page): Promise<Fixture> =>start: +(i * shot).toFixed(3),in: inPoint,out: +(inPoint + shot).toFixed(3),+ ...(dressed+ ? {+ look: "vhs",+ lookAmount: 0.8,+ grade: { contrast: 12, saturation: 10, temperature: 8, shadows: -6 },+ }+ : {}),})),audioClips: [],overlays: [],@@ -397,7 +424,7 @@ const seedMontage = async (page: Page): Promise<Fixture> =>duration: +(ins.length * shot).toFixed(3),};},- { ins: SHOT_INS, shot: SHOT_S, srcDuration: MONTAGE_SRC_S }+ { ins: SHOT_INS, shot: SHOT_S, srcDuration: MONTAGE_SRC_S, dressed });/** Where the scene-cut fixture changes scene: every three seconds. */@@ -798,11 +825,11 @@ const playbackCase = (name: string, transitions: boolean): EvalCase => ({* pass answers "does a fast cut play"; several answer "does it still play after* a while", which is the complaint this case was written for.*/-const montageCase = (name: string, passes: number): EvalCase => ({+const montageCase = (name: string, passes: number, dressed = false): EvalCase => ({name,bucket: "playback",transitions: false,- seed: seedMontage,+ seed: seedMontage(dressed),run: async (page, fx) => {await rewind(page);// One pass before the trace, so what follows measures playing rather than@@ -934,7 +961,7 @@ const montageCase = (name: string, passes: number): EvalCase => ({if (stallShare > GATE.stallShare) {notes.push(`picture hitched a quarter second behind on ${(stallShare * 100).toFixed(1)}% of frames`);}- if (passes > 1 && decay.last > Math.max(0.02, decay.first * GATE.decayRatio)) {+ if (passes > 1 && decay.last > Math.max(GATE.decayFloor, decay.first * GATE.decayRatio)) {notes.push(`late frames rose ${decay.first} → ${decay.last} over the play`);}const warmMb = Math.max(0, ...trace.warmMb);@@ -1307,6 +1334,9 @@ const CASES: EvalCase[] = [playbackCase("play-with-transitions", true),montageCase("play-fast-cuts", 1),montageCase("play-fast-cuts-long", 4),+ // The same montage graded and looked, which is what the cost of a real+ // project's frame looks like on a machine that has none to spare.+ montageCase("play-fast-cuts-graded", 2, true),idleCase,filmstripCase,filmstripCutsCase,