Checkpoint 2

generate-demuxer.ts

@@ -1,65 +1,88 @@
-const graph_density = 1000;
+import findLoudness, { SwapPoint } from "./find-loudness";
+
+const graph_density = 8000;
 
 const threshold_at_point = 1;
 
-const inertia_s = 0.1;
+const inertia_s = 0.3;
 const inertia_samples = inertia_s * graph_density;
 
-let position = 0;
+const s = (n: number) => n / graph_density;
-let last_swap_position = 0;
-
-const s = (n: number) => Math.round(n / graph_density);
 
 const minutes = (units: number) => Math.floor(s(units) / 60);
 
+const hours = (units: number) => Math.floor(units / graph_density / 60 / 60);
+
 const formatTime = (units: number) =>
-	`${minutes(units)}:${Math.floor(s(units) % 60)}`;
+	`${hours(units)}:${minutes(units)}:${Math.floor(s(units) % 60)}`;
 
-let keep_loud_until = 0;
+type Mode = { left: boolean; right: boolean };
 
-let total_speaking = 0;
+async function run() {
+	const [left_breaks, right_breaks] = await Promise.all([
+		findLoudness(
+			"/tmp/leftraw",
+			threshold_at_point,
+			inertia_samples,
+			"left"
+		),
+		findLoudness(
+			"/tmp/rightraw",
+			threshold_at_point,
+			inertia_samples,
+			"right"
+		),
+	]);
+
+	const merged = [...left_breaks, ...right_breaks].sort((a, b) =>
+		a.position_start < b.position_start
+			? -1
+			: a.position_start > b.position_start
+			? 1
+			: 0
+	);
 
-const results: [string, number][] = [];
+	// console.log("left breaks:", left_breaks);
+	// console.log(`right_breaks`, right_breaks);
+	// console.log(`merged`, merged);
 
-process.stdin.on("readable", () => {
+	function new_mode(m: Mode, s: SwapPoint): Mode {
-	let chunk: Buffer | null;
+		return { ...m, [s.label]: s.loud };
-	let was_loud_last_time = false;
-	while ((chunk = process.stdin.read()) !== null) {
-		for (let i = 0; i < chunk.byteLength; i++) {
-			position++;
-			const byte = chunk[i];
-			const volume = Math.abs(byte - 128);
-			const is_loud: boolean =
-				volume > threshold_at_point || position < keep_loud_until;
-			if (is_loud) {
-				total_speaking++;
 	}
-			if (is_loud != was_loud_last_time) {
+
-				results.push([
+	function mode_to_string(mode: Mode) {
-					is_loud ? "silence" : "speaking",
+		if (mode.left && mode.right) {
-					position - last_swap_position,
+			return "both";
-				]);
-				last_swap_position = position;
-				was_loud_last_time = is_loud;
 		}
-			if (volume > threshold_at_point) {
+		for (const side of ["left", "right"]) {
-				keep_loud_until = position + inertia_samples;
+			if (mode[side as keyof Mode]) {
+				return side;
 			}
 		}
+		return "none";
 	}
-});
-
-const mode_to_image = (mode: string) => {
-	return mode === "silence" ? "pics/cisza.png" : "pics/kuba.png";
-};
-
-process.stdin.on("end", () => {
-	results.forEach(([mode, duration_units]) => {
-		console.log("file", `'${mode_to_image(mode)}'`);
-		console.log("duration", (duration_units / graph_density).toFixed(4));
-	});
 
-	console.log("file", `'${mode_to_image(results[results.length - 1][0])}'`);
+	console.log("file", `${process.cwd()}/pics/none.png`);
-
+	let last_point = 0;
-	// console.log(formatTime(total_speaking), formatTime(position));
+	let mode: Mode = { left: false, right: false };
-});
+	let last_file;
+	let total = 0;
+	for (let i = 2; i < merged.length; i++) {
+		const point = merged[i];
+		mode = new_mode(mode, point);
+		const file = `${process.cwd()}/pics/${mode_to_string(mode)}.png`;
+		const duration = (point.position_start - last_point) / graph_density;
+		console.log(
+			"duration",
+			(point.position_start - last_point) / graph_density
+		);
+		console.log("file", file);
+		last_point = point.position_start;
+		last_file = file;
+		total += duration * graph_density;
+	}
+	console.log("duration", merged[merged.length - 1].duration / graph_density);
+	console.log("file", last_file);
+	console.error(total, formatTime(total));
+}
+run();

generate.sh

@@ -11,25 +11,27 @@ input=/home/kuba/Downloads/podcast-01-after-effects.mp3  # tutaj dajemy ścieżk
 aresample=8000   # to bez zmian
 
 echo dzielimy mp3 na dwa osobne wav
-ffmpeg -i $input -map_channel 0.0.0 /tmp/left.wav -map_channel 0.0.1 /tmp/right.wav
+#ffmpeg -i $input -map_channel 0.0.0 /tmp/left.wav -map_channel 0.0.1 /tmp/right.wav
 
 echo na dwóch wątkach generujemy surowe pliki
-ffmpeg -i /tmp/left.wav -ac 1 -filter:a aresample=$aresample -map 0:a -c:a pcm_u8 -f data -  > /tmp/leftraw &
+#ffmpeg -i /tmp/left.wav -ac 1 -filter:a aresample=$aresample -map 0:a -c:a pcm_u8 -f data -  > /tmp/leftraw &
-ffmpeg -i /tmp/right.wav -ac 1 -filter:a aresample=$aresample -map 0:a -c:a pcm_u8 -f data -  > /tmp/rightraw &
+#ffmpeg -i /tmp/right.wav -ac 1 -filter:a aresample=$aresample -map 0:a -c:a pcm_u8 -f data -  > /tmp/rightraw &
 
 # czekamy aż obydwa wątki się zakończą
-wait;
+#wait;
 
 echo "generating the demuxers...";
 
 # generuje ścieżki do złożenia przez ffmpega:
-ts-node generate-demuxer.ts  > /tmp/demuxer.txt
+ts-node generate-demuxer.ts  > out/demuxer.txt
 
-mkdir -f out
+mkdir -p out
 
 # używa demuxer.txt żeby skleić końcowe video z dźwiękiem:
 echo generowanie całości
-ffmpeg -y -f concat -i /tmp/demuxer.txt -r 30 -tune stillimage -vsync vfr -pix_fmt yuv420p out/video.mp4
+ffmpeg -y -f concat -safe 0 -i out/demuxer.txt -r 30 -tune stillimage -vsync vfr -pix_fmt yuv420p out/video.mp4
+#                   ^ daję safe 0 aby przyjmowało bezwzględne ścieżki
+
 
 echo łączenie video z dźwiękiem:
-ffmpeg  -i video.mp4 -i $input -ac 1 -tune stillimage out/video-and-audio.mp4
+ffmpeg  -i out/video.mp4 -i $input -ac 1 -tune stillimage out/video-and-audio.mp4