diff --git a/bytebeat_compiler.py b/bytebeat_compiler.py
index bd018b0..57d9097 100644
--- a/bytebeat_compiler.py
+++ b/bytebeat_compiler.py
@@ -124,6 +124,11 @@ if __name__ == "__main__":
 			 "default = +0 samples")
 	parser.add_argument("-a", "--no-return", default=False, action="store_true",
 		help="do not insert return statement before the code")
+	parser.add_argument("-u", "--mode", default="sequential", type=str,
+		help="mode of writing: `sequential` or `instant` (the latter is not "
+		     "recommended, since the whole result would be stored in RAM)")
+	parser.add_argument("-n", "--block-size", default=65536, type=int,
+		help="sequential mode only: block size of each sequence, bytes")
 	parser.add_argument("-q", "--silent", default=False, action="store_true",
 		help="do not output anything during generation")
 	parser.add_argument("-v", "--verbose", default=False, action="store_true",
@@ -186,6 +191,9 @@ if __name__ == "__main__":
 		print("CLI: Count of samples should be greater than zero.")
 		raise SystemExit
 
+	if args.mode != "sequential" and args.mode != "instant":
+		print("Invalid mode '%s'" % args.mode)
+
 	rewrite_file(PATHS["substitute"], substitute_vars({
 		"bytebeat_contents": bytebeat_contents,
 		"output_file": C_str_repr(args.output),
@@ -199,6 +207,8 @@ if __name__ == "__main__":
 		"fp_return_type": args.floating_point,
 		"channels": args.channels,
 		"length": samples,
+		"sequential_mode": args.mode == "sequential",
+		"block_size": args.block_size,
 		"silent_mode": args.silent,
 		"verbose_mode": args.verbose and not args.silent,
 		"fwrite_le": "../" + PATHS["fwrite_le_header"]
diff --git a/src/template.c b/src/template.c
index 07ffe37..0b88c7a 100644
--- a/src/template.c
+++ b/src/template.c
@@ -41,6 +41,9 @@ const char* ANSI_CLEAR = __ANSI_CLEAR_STRING;
 #define PRODUCT (LENGTH * CHANNELS)
 #define FREQUENCY_OF_STATUS_REPORTING 5000
 
+#define SEQUENTIAL_MODE `sequential_mode`
+#define BLOCK_SIZE_BYTES `block_size`
+
 #define FP_RETURN_TYPE `fp_return_type`
 #define PRECALCULATED_RATIO `precalculated_ratio`
 
@@ -71,8 +74,12 @@ bytebeat(long double w)
 {
 #if PRECALCULATED_RATIO
 	`final_sample_rate_code`
-#else
+#elif ORIGINAL_SAMPLE_RATE != SAMPLE_RATE
+# if SAMPLE_RATE > ORIGINAL_SAMPLE_RATE
 	w /= ((long double) SAMPLE_RATE) / ((long double) ORIGINAL_SAMPLE_RATE);
+# else
+	w *= ((long double) ORIGINAL_SAMPLE_RATE / ((long double) SAMPLE_RATE));
+# endif
 #endif
 	uintmax_t t = (uintmax_t) w;
 	`bytebeat_contents`;
@@ -122,7 +129,7 @@ main(void)
 		printf("%" PRIuMAX " samples", samples);
 
 	printf(
-#if VERBOSE_MODE
+#if VERBOSE_MODE || SEQUENTIAL_MODE
 		"\n"
 #endif
 		"\n"
@@ -132,6 +139,11 @@ main(void)
 #endif
 
 	// * write WAVE headers
+	// 0. log
+#if SEQUENTIAL_MODE
+	printf("Writing WAVE headers...\n");
+#endif
+
 	// 1. open file
 	FILE* output_file = fopen(OUTPUT_FILE, "wb");
 	if (output_file == NULL) {
@@ -146,10 +158,10 @@ main(void)
 			5 * 4   /* 4 uint32_t values */         +
 			4 * 2   /* 5 uint16_t values */         +
 			PRODUCT /* sample data */,
-		fmt_data_length = 16 /*
-		                      * length of format data before this value
-		                      * in the file format structure
-				      */,
+		fmt_data_length = 16 /* <--
+		 * length of format data before this value
+		 * in the file format structure
+		 */,
 		sample_rate = SAMPLE_RATE,
 		byte_rate = (SAMPLE_RATE * BIT_DEPTH * CHANNELS) / 8;
 	uint16_t fmt_type = 1, // format type is PCM
@@ -174,93 +186,155 @@ main(void)
 	fwrite_le(&buffer_size,     4, 1, output_file); // L : Subchunk2Size : 4
 
 	// 4. write sample data
-	// * allocate heap for sample data
-	SAMPLE_TYPE* buffer = calloc(PRODUCT, sizeof(SAMPLE_TYPE));
-
-	if (buffer == NULL) {
-		perror("calloc");
-		return 1;
+#if SEQUENTIAL_MODE
+	const size_t BLOCK_SIZE = BLOCK_SIZE_BYTES / (sizeof(SAMPLE_TYPE) /
+		sizeof(uint8_t));
+	if (BLOCK_SIZE < 1) {
+		printf("The block size %" PRIuMAX " is too small, should be at least "
+		       "%" PRIuMAX " bytes\n", (uintmax_t) BLOCK_SIZE_BYTES,
+		       (uintmax_t) (sizeof(SAMPLE_TYPE) / sizeof(uint8_t)));
+		exit(EXIT_FAILURE);
 	}
+	const size_t MAX = (PRODUCT + (BLOCK_SIZE - 1)) / BLOCK_SIZE;
 
-	// * bytebeat generating loop
-	const uintmax_t product_minus_1 = PRODUCT - 1,
-		bit_depth_limiter = BIT_DEPTH_LIMITER
-#if FP_RETURN_TYPE
-			+ 1
-#endif
-#if BIT_DEPTH < 8
-		,
-		bit_depth_stretch = ((long double) BIT_DEPTH) / 8.L
-#endif
-		;
-
-	for (size_t w = 0; w < PRODUCT; w++) {
-		// 1. generate audio data
-#if FP_RETURN_TYPE
-		long double bytebeat_res = floor(bytebeat((long double) floor(w)));
-#elif IS_SIGNED
-		intmax_t bytebeat_res = (intmax_t) bytebeat(floor((long double) w));
+	size_t w = 0;
+	for (size_t seq = 0; seq < MAX; seq++) {
 #else
-		uintmax_t bytebeat_res = (uintmax_t) bytebeat(floor((long double) w));
+	do {
 #endif
+		size_t calc_block_size = BLOCK_SIZE;
+		if ((w + BLOCK_SIZE) >= PRODUCT)
+			calc_block_size = PRODUCT - w;
 
-		// 2. if signed, then wrap up into unsigned
-#if IS_SIGNED
-		bytebeat_res = signed_to_unsigned(bytebeat_res);
-#endif
-
-		// 3. convert audio data to sample
-		SAMPLE_TYPE sample_res = (SAMPLE_TYPE)
-#if FP_RETURN_TYPE
-			fmodl(bytebeat_res, bit_depth_limiter);
+		// * allocate heap for sample data
+		SAMPLE_TYPE* buffer = calloc(
+#if SEQUENTIAL_MODE
+			calc_block_size,
 #else
-			((uintmax_t) bytebeat_res & bit_depth_limiter);
+			PRODUCT,
 #endif
+			sizeof(SAMPLE_TYPE));
 
-		// 4. if bit depth is less than 8, stretch it
-#if BIT_DEPTH < 8
-		sample_res = (SAMPLE_TYPE)
-			((long double) sample_res * bit_depth_stretch);
-#endif
-
-		// 5. save sample into buffer
-		buffer[w] = sample_res;
-
-		// 6. log
-#if VERBOSE_MODE
-		if (w % FREQUENCY_OF_STATUS_REPORTING == 0 ||
-			w >= product_minus_1 /* or if writing last sample */) {
-			printf(
-			  "%sremaining samples = %18" PRIuMAX " (%.2Lf%% done)",
-			  ANSI_CLEAR,
-			  product_minus_1 - w,
-			  ((long double) w * 100) / (long double) PRODUCT);
-			fflush(stdout);
+		if (buffer == NULL) {
+			perror("calloc");
+			return 1;
 		}
+
+		// * bytebeat generating loop
+		const uintmax_t product_minus_1 = PRODUCT - 1,
+			bit_depth_limiter = BIT_DEPTH_LIMITER
+#if FP_RETURN_TYPE
+				+ 1
 #endif
-	}
+#if BIT_DEPTH < 8
+			,
+			bit_depth_stretch = ((long double) BIT_DEPTH) / 8.L
+#endif
+			;
+
+#if SEQUENTIAL_MODE
+		size_t idx = 0;
+		for (; idx < BLOCK_SIZE && w < PRODUCT; idx++, w++) {
+#else
+		for (size_t w = 0; w < PRODUCT; w++) {
+#endif
+			// 1. generate audio data
+#if FP_RETURN_TYPE
+			long double bytebeat_res = floor(bytebeat((long double) floor(w)));
+#elif IS_SIGNED
+			intmax_t bytebeat_res = (intmax_t) bytebeat(floor((long double) w));
+#else
+			uintmax_t bytebeat_res = (uintmax_t) bytebeat(floor((long double) w));
+#endif
+
+			// 2. if signed, then wrap up into unsigned
+#if IS_SIGNED
+			bytebeat_res = signed_to_unsigned(bytebeat_res);
+#endif
+
+			// 3. convert audio data to sample
+			SAMPLE_TYPE sample_res = (SAMPLE_TYPE)
+#if FP_RETURN_TYPE
+				fmodl(bytebeat_res, bit_depth_limiter);
+#else
+				((uintmax_t) bytebeat_res & bit_depth_limiter);
+#endif
+
+			// 4. if bit depth is less than 8, stretch it
+#if BIT_DEPTH < 8
+			sample_res = (SAMPLE_TYPE)
+				((long double) sample_res * bit_depth_stretch);
+#endif
+
+			// 5. save sample into buffer
+#if SEQUENTIAL_MODE
+			buffer[idx] = sample_res;
+#else
+			buffer[w] = sample_res;
+#endif
+
+			// 6. log
+#if VERBOSE_MODE
+			if (w % FREQUENCY_OF_STATUS_REPORTING == 0 ||
+				w >= product_minus_1 /* or if writing last sample */) {
+				printf(
+				  "%sremaining samples = %18" PRIuMAX " (%.2Lf%% done)",
+				  ANSI_CLEAR,
+				  product_minus_1 - w,
+				  ((long double) w * 100) / (long double) PRODUCT);
+				fflush(stdout);
+			}
+#endif
+		}
 
 #if !SILENT_MODE
-	printf("%s", ANSI_CLEAR);
+		printf("%s", ANSI_CLEAR);
 
-	// 5. log
-	printf("\nWriting out file " OUTPUT_FILE "...\n");
-	fflush(stdout);
+		// 5. log
+#if SEQUENTIAL_MODE
+		printf("Writing out file " OUTPUT_FILE " (part %" PRIuMAX "/%"
+		       PRIuMAX ", %.2Lf%%)...\n", (uintmax_t) seq + 1,
+		       (uintmax_t) MAX,
+		       ((long double) seq * 100.L) / ((long double) MAX));
+#else
+		printf("\nWriting out file " OUTPUT_FILE "...\n");
+#endif
+		fflush(stdout);
 #endif
 
-	// * save the sample data into the file
-	fwrite_le(buffer, sizeof(SAMPLE_TYPE), buffer_size, output_file);
+		// * save the sample data into the file
+		fwrite_le(buffer, sizeof(SAMPLE_TYPE),
+#if SEQUENTIAL_MODE
+			calc_block_size,
+#else
+			PRODUCT,
+#endif
+			output_file);
+#if SEQUENTIAL_MODE
+		fflush(output_file);
+#endif
+
+#if SEQUENTIAL_MODE
+		// * free allocated heap
+		free(buffer);
+#endif
+	}
+#if !SEQUENTIAL_MODE
+		while (0);
+#endif
 
 	// 6. close file
 	fclose(output_file);
 
+#if !SEQUENTIAL_MODE
+	// * free allocated heap
+	free(buffer);
+#endif
+
 	// * end of program
 #if !SILENT_MODE
 	printf("Done!\n");
 #endif
 
-	// * free allocated heap
-	free(buffer);
-
 	return 0;
 }