diff --git a/adapters/standalone.cpp b/adapters/standalone.cpp
index a5092ce9..09017459 100644
--- a/adapters/standalone.cpp
+++ b/adapters/standalone.cpp
@@ -113,7 +113,7 @@ int main(int argc, char* argv[]) {
 	logger::init();
 	random::init();
 
-	// Test stuff
+	// Test code
 	// exit(0);
 
 	// We can now install a signal handler and log the output
diff --git a/include/random.hpp b/include/random.hpp
index 2ba232cd..1fc9b732 100644
--- a/include/random.hpp
+++ b/include/random.hpp
@@ -1,5 +1,7 @@
 #pragma once
 #include <common.hpp>
+#include <random>
+#include <vector>
 
 
 namespace rack {
@@ -10,18 +12,105 @@ namespace rack {
 namespace random {
 
 
+/** xoroshiro128+. Very fast, not-cryptographic random number generator.
+From https://prng.di.unimi.it/
+Example:
+
+	std::random_device rd;
+	random::Xoroshiro128Plus rng(rd());
+	uint64_t r = rng();
+	uint32_t r = rng.u32();
+
+	std::uniform_real_distribution<float> uniform(0.f, 1.f);
+	float r = uniform(rng);
+
+	std::normal_distribution<> normal(0.0, 1.0);
+	double r = normal(rng);
+*/
+struct Xoroshiro128Plus {
+	uint64_t state[2];
+
+	Xoroshiro128Plus(uint64_t s0 = 1, uint64_t s1 = 0) {
+		seed(s0, s1);
+	}
+	void seed(uint64_t s0 = 1, uint64_t s1 = 0) {
+		state[0] = s0;
+		state[1] = s1;
+		operator()();
+	}
+
+	static uint64_t rotl(const uint64_t x, int k) {
+		return (x << k) | (x >> (64 - k));
+	}
+
+	uint64_t operator()() {
+		const uint64_t s0 = state[0];
+		uint64_t s1 = state[1];
+		const uint64_t result = s0 + s1;
+
+		s1 ^= s0;
+		state[0] = rotl(s0, 55) ^ s1 ^ (s1 << 14);
+		state[1] = rotl(s1, 36);
+
+		return result;
+	}
+	constexpr uint64_t min() {
+		return 0;
+	}
+	constexpr uint64_t max() {
+		return UINT64_MAX;
+	}
+
+	uint64_t u64() {
+		return operator()();
+	}
+	uint64_t u32() {
+		// Take top 32 bits which has better randomness properties.
+		return u64() >> 32;
+	}
+	uint16_t u16() {
+		return u64() >> 48;
+	}
+	uint8_t u8() {
+		return u64() >> 56;
+	}
+	float f32() {
+		// The multiplier is 2f7fffff in hex. This gives maximum precision of uint32_t -> float conversion and its image is [0, 1).
+		return u32() * 2.32830629e-10f;
+	}
+	float f64() {
+		return u64() * 5.421010862427522e-20;
+	}
+};
+
+
+// Easy random API
+
+extern thread_local Xoroshiro128Plus rng;
+
+
 /** Initializes the thread-local RNG state.
 Must call per-thread, otherwise the RNG will always return 0.
 */
 void init();
-/** Returns a uniform random uint32_t from 0 to UINT32_MAX */
-uint32_t u32();
 /** Returns a uniform random uint64_t from 0 to UINT64_MAX */
-uint64_t u64();
+inline uint64_t u64() {
+	return rng.u64();
+}
+/** Returns a uniform random uint32_t from 0 to UINT32_MAX */
+inline uint32_t u32() {
+	return rng.u32();
+}
 /** Returns a uniform random float in the interval [0.0, 1.0) */
-float uniform();
+inline float uniform() {
+	return rng.f32();
+}
 /** Returns a normal random number with mean 0 and standard deviation 1 */
 float normal();
+/** Fills an array with random bytes. */
+void buffer(uint8_t* out, size_t len);
+/** Creates a vector of random bytes. */
+std::vector<uint8_t> vector(size_t len);
 
 
 } // namespace random
diff --git a/src/random.cpp b/src/random.cpp
index 351b3bc2..08961052 100644
--- a/src/random.cpp
+++ b/src/random.cpp
@@ -11,55 +11,20 @@ namespace rack {
 namespace random {
 
 
-// xoroshiro128+
-// from http://xoroshiro.di.unimi.it/xoroshiro128plus.c
-
-thread_local uint64_t xoroshiro128plus_state[2];
+thread_local Xoroshiro128Plus rng;
 static std::atomic<uint64_t> threadCounter {0};
 
-static uint64_t rotl(const uint64_t x, int k) {
-	return (x << k) | (x >> (64 - k));
-}
-
-static uint64_t xoroshiro128plus_next(void) {
-	const uint64_t s0 = xoroshiro128plus_state[0];
-	uint64_t s1 = xoroshiro128plus_state[1];
-	const uint64_t result = s0 + s1;
-
-	s1 ^= s0;
-	xoroshiro128plus_state[0] = rotl(s0, 55) ^ s1 ^ (s1 << 14); // a, b
-	xoroshiro128plus_state[1] = rotl(s1, 36); // c
-
-	return result;
-}
 
 void init() {
-	// Do nothing if already initialized
-	if (xoroshiro128plus_state[0] || xoroshiro128plus_state[1])
-		return;
-
 	struct timeval tv;
 	gettimeofday(&tv, NULL);
-	xoroshiro128plus_state[0] = uint64_t(tv.tv_sec) * 1000000 + tv.tv_usec;
-	xoroshiro128plus_state[1] = threadCounter++;
-	// Shift a few times to fix the fact that the seed is not a uniform u64
-	for (int i = 0; i < 10; i++) {
-		xoroshiro128plus_next();
+	rng = Xoroshiro128Plus(uint64_t(tv.tv_sec) * 1000000 + tv.tv_usec, threadCounter++);
+	// Shift state a few times due to low seed entropy
+	for (int i = 0; i < 4; i++) {
+		rng();
 	}
 }
 
-uint32_t u32() {
-	return xoroshiro128plus_next() >> 32;
-}
-
-uint64_t u64() {
-	return xoroshiro128plus_next();
-}
-
-float uniform() {
-	// The multiplier is 2f7fffff in hex. This gives maximum precision of uint32_t -> float conversion and its image is [0, 1).
-	return u32() * 2.32830629e-10f;
-}
 
 float normal() {
 	// Box-Muller transform
@@ -77,5 +42,26 @@ float normal() {
 }
 
 
+void buffer(uint8_t* out, size_t len) {
+	for (size_t i = 0; i < len; i += 4) {
+		uint64_t r = u64();
+		out[i] = r;
+		if (i + 1 < len)
+			out[i + 1] = r >> 8;
+		if (i + 2 < len)
+			out[i + 2] = r >> 16;
+		if (i + 3 < len)
+			out[i + 3] = r >> 24;
+	}
+}
+
+
+std::vector<uint8_t> vector(size_t len) {
+	std::vector<uint8_t> v(len);
+	buffer(v.data(), len);
+	return v;
+}
+
+
 } // namespace random
 } // namespace rack