Fix MP4 recorder timestamp monotonicity

src/record/mp4_record_writer.cpp

@@ -81,6 +81,23 @@ AVRational frame_rate_rational(const float fps) {
 	return AVRational{scaled, 1000};
 }
 
+[[nodiscard]]
+std::uint64_t frame_period_ns(const AVRational frame_rate) {
+	if (frame_rate.num <= 0 || frame_rate.den <= 0) {
+		return 33'333'333ull;
+	}
+
+	const auto numerator =
+		static_cast<std::uint64_t>(frame_rate.den) * kNanosPerSecond;
+	const auto denominator = static_cast<std::uint64_t>(frame_rate.num);
+	if (denominator == 0) {
+		return 33'333'333ull;
+	}
+
+	const auto interval = numerator / denominator;
+	return interval == 0 ? 1ull : interval;
+}
+
 [[nodiscard]]
 std::vector<EncoderCandidate> encoder_candidates(const CodecType codec, const EncoderDeviceType device) {
 	const std::string hardware_name = codec == CodecType::H265 ? "hevc_nvenc" : "h264_nvenc";
@@ -303,6 +320,8 @@ struct Mp4RecordWriter::Impl {
 
 		codec = codec_arg;
 		frame_rate = frame_rate_rational(fps);
+		frame_period = frame_period_ns(frame_rate);
+		last_frame_pts_ns.reset();
 		input_pixel_format = input_pixel_format_arg;
 		source_pixel_format = av_pixel_format(input_pixel_format);
 		if (source_pixel_format == AV_PIX_FMT_NONE) {
@@ -439,7 +458,13 @@ struct Mp4RecordWriter::Impl {
 			frame->data,
 			frame->linesize);
 
-		frame->pts = static_cast<std::int64_t>(relative_timestamp_ns);
+		auto normalized_timestamp_ns = relative_timestamp_ns;
+		if (last_frame_pts_ns && normalized_timestamp_ns <= *last_frame_pts_ns) {
+			normalized_timestamp_ns = *last_frame_pts_ns + frame_period;
+		}
+
+		frame->pts = static_cast<std::int64_t>(normalized_timestamp_ns);
+		last_frame_pts_ns = normalized_timestamp_ns;
 
 		const auto send_result = avcodec_send_frame(encoder_context, frame);
 		if (send_result < 0) {
@@ -534,6 +559,8 @@ struct Mp4RecordWriter::Impl {
 		encoder_name.clear();
 		using_hardware = false;
 		trailer_written = false;
+		frame_period = 33'333'333ull;
+		last_frame_pts_ns.reset();
 		resolved_settings = ResolvedEncoderSettings{};
 	}
 
@@ -551,6 +578,8 @@ struct Mp4RecordWriter::Impl {
 	AVPixelFormat encoder_pixel_format{AV_PIX_FMT_NONE};
 	AVPixelFormat source_pixel_format{AV_PIX_FMT_NONE};
 	AVRational frame_rate{30, 1};
+	std::uint64_t frame_period{33'333'333ull};
+	std::optional<std::uint64_t> last_frame_pts_ns{};
 	std::string encoder_name{};
 	ResolvedEncoderSettings resolved_settings{};
 	bool using_hardware{false};

src/tools/zed_svo_mp4_support.cpp

@@ -64,6 +64,23 @@ AVRational frame_rate_rational(const float fps) {
 	return AVRational{scaled, 1000};
 }
 
+[[nodiscard]]
+std::uint64_t frame_period_ns(const AVRational frame_rate) {
+	if (frame_rate.num <= 0 || frame_rate.den <= 0) {
+		return 33'333'333ull;
+	}
+
+	const auto numerator =
+		static_cast<std::uint64_t>(frame_rate.den) * 1'000'000'000ull;
+	const auto denominator = static_cast<std::uint64_t>(frame_rate.num);
+	if (denominator == 0) {
+		return 33'333'333ull;
+	}
+
+	const auto interval = numerator / denominator;
+	return interval == 0 ? 1ull : interval;
+}
+
 [[nodiscard]]
 std::vector<EncoderCandidate> encoder_candidates(const CodecType codec, const EncoderDeviceType device) {
 	const std::string hardware_name = codec == CodecType::H265 ? "hevc_nvenc" : "h264_nvenc";
@@ -314,6 +331,8 @@ struct Mp4Writer::Impl {
 
 		codec = codec_arg;
 		frame_rate = frame_rate_rational(fps);
+		frame_period = frame_period_ns(frame_rate);
+		last_frame_pts_ns.reset();
 		auto encoder = open_encoder(codec, encoder_device, width, height, frame_rate, tuning);
 		if (!encoder) {
 			return std::unexpected(encoder.error());
@@ -447,7 +466,13 @@ struct Mp4Writer::Impl {
 			frame->data,
 			frame->linesize);
 
-		frame->pts = static_cast<std::int64_t>(relative_timestamp_ns);
+		auto normalized_timestamp_ns = relative_timestamp_ns;
+		if (last_frame_pts_ns && normalized_timestamp_ns <= *last_frame_pts_ns) {
+			normalized_timestamp_ns = *last_frame_pts_ns + frame_period;
+		}
+
+		frame->pts = static_cast<std::int64_t>(normalized_timestamp_ns);
+		last_frame_pts_ns = normalized_timestamp_ns;
 
 		const auto send_result = avcodec_send_frame(encoder_context, frame);
 		if (send_result < 0) {
@@ -542,6 +567,8 @@ struct Mp4Writer::Impl {
 		encoder_name.clear();
 		using_hardware = false;
 		trailer_written = false;
+		frame_period = 33'333'333ull;
+		last_frame_pts_ns.reset();
 		resolved_settings = ResolvedEncoderSettings{};
 	}
 
@@ -558,6 +585,8 @@ struct Mp4Writer::Impl {
 	SwsContext *scaler{nullptr};
 	AVPixelFormat encoder_pixel_format{AV_PIX_FMT_NONE};
 	AVRational frame_rate{30, 1};
+	std::uint64_t frame_period{33'333'333ull};
+	std::optional<std::uint64_t> last_frame_pts_ns{};
 	std::string encoder_name{};
 	ResolvedEncoderSettings resolved_settings{};
 	bool using_hardware{false};
@@ -636,13 +665,6 @@ std::string_view tune_name(const TuneKind tune) {
 	return "low-latency";
 }
 
-std::uint64_t frame_period_ns(const float fps) {
-	if (!(fps > 0.0f)) {
-		return 33'333'333ull;
-	}
-	return static_cast<std::uint64_t>(std::llround(1'000'000'000.0 / static_cast<double>(fps)));
-}
-
 std::filesystem::path derive_output_path(const std::filesystem::path &input_path) {
 	auto output_path = input_path;
 	output_path.replace_extension(".mp4");