diff --git a/CMakeLists.txt b/CMakeLists.txt
index 8a9216f..269e810 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -19,10 +19,13 @@ add_subdirectory(third-party/physfs)
 
 set(SOURCE_FILES
         third-party/physfs/extras/physfsrwops.c
+        third-party/stb/stb_vorbis.c
+
         src/config.h
         src/context.h
         src/context.c
         src/main.c
+        src/audio.c
         src/util.c src/util.h
         src/rendering.c src/rendering.h
         src/textures.c src/textures.h
@@ -75,8 +78,7 @@ else()
                 -Wshadow
                 -Wdouble-promotion
                 -Wconversion -Wno-sign-conversion
-                -Werror=vla
-                -Werror=alloca)
+                -Werror=vla)
         set(BUILD_FLAGS
                 # these SHOULDN'T break anything...
                 -fno-math-errno
@@ -88,6 +90,7 @@ else()
                 -flto)
         set(BUILD_FLAGS_DEBUG
                 -g3
+                -gdwarf
                 -fsanitize-trap=undefined)
         target_compile_options(${PROJECT_NAME} PRIVATE
                                                ${WARNING_FLAGS}
diff --git a/data/music/test.ogg b/data/music/test.ogg
new file mode 100644
index 0000000..10bbd82
--- /dev/null
+++ b/data/music/test.ogg
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a964aeb4120f1542d7ff3371eb640a67354a082f44e71a58020c9002163e361
+size 2127638
diff --git a/src/audio.c b/src/audio.c
new file mode 100644
index 0000000..85848fb
--- /dev/null
+++ b/src/audio.c
@@ -0,0 +1,228 @@
+#include "private/audio.h"
+#include "audio.h"
+#include "context.h"
+#include "util.h"
+
+#include <SDL2/SDL.h>
+#include <stb_ds.h>
+#include <physfs.h>
+
+#define STB_VORBIS_HEADER_ONLY
+#include <stb_vorbis.c>
+
+#include <stdint.h>
+#include <string.h>
+
+/* TODO: default to float sampling format? */
+
+static const char *audio_exts[audio_file_type_count] = {
+    ".ogg", /* audio_file_type_ogg */
+    ".xm",  /* audio_file_type_xm */
+};
+
+
+/* TODO: count frames without use, free the memory when threshold is met */
+/* stores path to data hash, useful for sound effects */
+static struct audio_file_cache {
+    char *key;
+    struct audio_file_cache_value {
+        unsigned char *data;
+        size_t len;
+    } value;
+} *audio_file_cache;
+
+
+static int64_t get_audio_data(const char *path, unsigned char **data) {
+    const struct audio_file_cache *cache = shgetp_null(audio_file_cache, path);
+    if (!cache) {
+        unsigned char *file;
+        int64_t len = file_to_bytes(path, &file);
+        if (len == -1) {
+            CRY("Audio error", "Error reading file");
+            return -1;
+        }
+
+        const struct audio_file_cache_value value = { file, (size_t)len };
+        shput(audio_file_cache, path, value);
+
+        *data = file;
+        return len;
+    }
+
+    *data = cache->value.data;
+    return (int64_t)cache->value.len;
+}
+
+
+void play_audio(const char *path, const char *channel) {
+    const struct audio_channel_pair *pair = shgetp_null(ctx.audio_channels, channel);
+    if (!pair)
+        play_audio_ex(path, channel, get_default_audio_args());
+    else
+        play_audio_ex(path, channel, pair->value.args);
+}
+
+
+static t_audio_file_type infer_audio_file_type(const char *path) {
+    size_t path_len = strlen(path);
+
+    for (int i = 0; i < audio_file_type_count; ++i) {
+        size_t ext_length = strlen(audio_exts[i]);
+        if (path_len <= ext_length)
+            continue;
+
+        if (strcmp(&path[path_len - ext_length], audio_exts[i]) == 0)
+            return (t_audio_file_type)i;
+    }
+
+    return audio_file_type_unknown;
+}
+
+
+/* TODO: error propagation and clearing of resources on partial success? */
+/*       or should we expect things to simply fail? */
+static union audio_context init_audio_context(const char *path, t_audio_file_type type) {
+    switch (type) {
+    case audio_file_type_ogg: {
+        unsigned char *data;
+        int64_t len = get_audio_data(path, &data);
+        if (len == -1) {
+            CRY("Audio error", "Error reading file");
+            break;
+        }
+
+        int error = 0;
+        stb_vorbis* handle = stb_vorbis_open_memory(data, (int)len, &error, NULL);
+        if (error != 0) {
+            CRY("Audio error", "Error reading .ogg file");
+            break;
+        }
+
+        stb_vorbis_info info = stb_vorbis_get_info(handle);
+
+        return (union audio_context){
+            .vorbis = {
+                .data = data,
+                .handle = handle,
+                .frequency = info.sample_rate,
+                .channel_count = (uint8_t)info.channels,
+            }
+        };
+    }
+
+    default:
+        CRY("Audio error", "Unhandled audio format (in init)");
+        return (union audio_context){0};
+    }
+
+    return (union audio_context){0};
+}
+
+
+static void repeat_audio(struct audio_channel *channel) {
+    (void)channel;
+
+    /* TODO */
+}
+
+
+void play_audio_ex(const char *path, const char *channel, t_play_audio_args args) {
+    struct audio_channel_pair *pair = shgetp_null(ctx.audio_channels, channel);
+
+    /* create a channel if it doesn't exist */
+    if (!pair) {
+        t_audio_file_type file_type = infer_audio_file_type(path);
+        struct audio_channel new_channel = {
+            .args = args,
+            .file_type = file_type,
+            .context = init_audio_context(path, file_type),
+            .path = path,
+        };
+        shput(ctx.audio_channels, channel, new_channel);
+        pair = shgetp_null(ctx.audio_channels, channel);
+    }
+
+    /* TODO: destroy and create new context when channel is reused for different file */
+
+    /* works for both restarts and new audio */
+    if (strcmp(pair->value.path, path) == 0)
+        repeat_audio(&pair->value);
+}
+
+
+t_play_audio_args get_default_audio_args(void) {
+    return (t_play_audio_args){
+        .repeat = false,
+        .crossfade = false,
+        .volume = 1.0f,
+        .panning = 0.0f,
+    };
+}
+
+
+static void audio_sample_and_mixin_channel(struct audio_channel *channel,
+                                           uint8_t *stream,
+                                           int len) {
+    int16_t *const sstream = (int16_t *)stream;
+    static uint8_t buffer[16384];
+    const int buffer_frames = sizeof (buffer) / sizeof (int16_t);
+
+    switch (channel->file_type) {
+    case audio_file_type_ogg: {
+        /* feed stream for needed conversions */
+        for (int i = 0; i < (len / 2);) {
+            const int n_frames = ((len / 2) - i) > buffer_frames ? buffer_frames : (len / 2) - i;
+
+            /* TODO: handle end of file */
+            const int samples_per_channel = stb_vorbis_get_samples_short_interleaved(
+                channel->context.vorbis.handle,
+                channel->context.vorbis.channel_count,
+                (int16_t *)buffer,
+                n_frames);
+
+            /* panning and mixing */
+
+#if AUDIO_N_CHANNELS == 2
+
+            for (int s = 0; s < samples_per_channel * ctx.audio_stream_channel_count; s += 2) {
+                /* left channel */
+                {
+                    const float panning = fminf(fabsf(channel->args.panning - 1.0f), 1.0f);
+                    const float volume = channel->args.volume * panning;
+                    sstream[i + s] += (int16_t)(((int16_t *)buffer)[s] * volume);
+                }
+
+                /* right channel */
+                {
+                    const float panning = fminf(fabsf(channel->args.panning + 1.0f), 1.0f);
+                    const float volume = channel->args.volume * panning;
+                    sstream[i + s + 1] += (int16_t)(((int16_t *)buffer)[s + 1] * volume);
+                }
+            }
+#else
+#error "Unimplemented channel count"
+#endif
+
+            i += samples_per_channel * ctx.audio_stream_channel_count;
+        }
+
+        break;
+    }
+
+    default:
+        CRY("Audio error", "Unhandled audio format (in sampling)");
+        break;
+    }
+}
+
+
+void audio_callback(void *userdata, uint8_t *stream, int len) {
+    (void)userdata;
+
+    /* prepare for mixing */
+    SDL_memset(stream, sizeof (uint16_t), len * ctx.audio_stream_channel_count);
+
+    for (int i = 0; i < shlen(ctx.audio_channels); ++i) {
+        audio_sample_and_mixin_channel(&ctx.audio_channels[i].value, stream, len);
+    }
+}
diff --git a/src/audio.h b/src/audio.h
new file mode 100644
index 0000000..5812b2e
--- /dev/null
+++ b/src/audio.h
@@ -0,0 +1,36 @@
+#ifndef AUDIO_H
+#define AUDIO_H
+
+#include <stdbool.h>
+
+
+typedef struct play_audio_args {
+    /* default: false */
+    bool repeat;
+    /* crossfade between already playing audio on a given channel, if any */
+    /* default: false */
+    bool crossfade;
+    /* range: 0.0f to 1.0f */
+    /* default: 100.0f */
+    float volume;
+    /* range: -1.0 to 1.0f */
+    /* default: 0.0f */
+    float panning;
+} t_play_audio_args;
+
+
+/* plays audio file at specified channel or anywhere if NULL is passed */
+/* path must contain valid file extension to infer which file format it is */
+/* supported formats: .ogg, .xm */
+/* preserves args that are already specified on the channel */
+void play_audio(const char *path, const char *channel);
+
+void play_audio_ex(const char *path, const char *channel, t_play_audio_args args);
+
+void set_audio_args(const char *channel, t_play_audio_args args);
+
+t_play_audio_args get_audio_args(const char *channel);
+
+t_play_audio_args get_default_audio_args(void);
+
+#endif
diff --git a/src/config.h b/src/config.h
index c1d6e95..805214f 100644
--- a/src/config.h
+++ b/src/config.h
@@ -24,6 +24,9 @@
 
 #define NUM_KEYBIND_SLOTS 8
 
+#define AUDIO_FREQUENCY 48000
+#define AUDIO_N_CHANNELS 2
+
 /* 1024 * 1024 */
 /* #define UMKA_STACK_SIZE 1048576 */
 
diff --git a/src/context.c b/src/context.c
index 45cd133..2c8a7bc 100644
--- a/src/context.c
+++ b/src/context.c
@@ -1,3 +1,3 @@
 #include "context.h"
 
-t_ctx ctx;
+t_ctx ctx = {0};
diff --git a/src/context.h b/src/context.h
index 55d4d8f..5a49ca1 100644
--- a/src/context.h
+++ b/src/context.h
@@ -26,7 +26,11 @@ typedef struct context {
     struct rect_primitive *render_queue_rectangles;
     struct circle_primitive *render_queue_circles;
 
-    struct audio_channel *audio_channels;
+    struct audio_channel_pair *audio_channels;
+    SDL_AudioDeviceID audio_device;
+    int audio_stream_frequency;
+    SDL_AudioFormat audio_stream_format;
+    uint8_t audio_stream_channel_count;
 
     struct circle_radius_cache_item *circle_radius_hash;
 
diff --git a/src/game/scenes/ingame.c b/src/game/scenes/ingame.c
index 24b9496..749f3a3 100644
--- a/src/game/scenes/ingame.c
+++ b/src/game/scenes/ingame.c
@@ -2,6 +2,8 @@
 #include "title.h"
 #include "scene.h"
 
+#include "../../audio.h"
+
 
 static void ingame_tick(struct state *state) {
     struct scene_ingame *scn = (struct scene_ingame *)state->scene;
@@ -28,5 +30,10 @@ struct scene *ingame_scene(struct state *state) {
     new_scene->world = world_create();
     new_scene->player = player_create(new_scene->world);
 
+    play_audio_ex("music/test.ogg", "soundtrack", (t_play_audio_args){
+        .volume = 0.8f,
+        .panning = -0.5f
+    });
+
     return (struct scene *)new_scene;
 }
diff --git a/src/main.c b/src/main.c
index 8804245..ad1b544 100644
--- a/src/main.c
+++ b/src/main.c
@@ -4,10 +4,11 @@
 #include "util.h"
 #include "textures.h"
 #include "game/game.h"
+#include "private/audio.h"
 
 #include <SDL2/SDL.h>
-#include <SDL2/SDL_image.h>
 #include <SDL2/SDL_ttf.h>
+#include <SDL2/SDL_image.h>
 #include <physfs.h>
 #include <stb_ds.h>
 
@@ -179,6 +180,26 @@ static bool initialize(void) {
     SDL_RenderSetLogicalSize(ctx.renderer, RENDER_BASE_WIDTH, RENDER_BASE_HEIGHT);
     SDL_GetRendererOutputSize(ctx.renderer, &ctx.window_w, &ctx.window_h);
 
+    /* audio initialization */
+    {
+
+    SDL_AudioSpec request, got;
+    SDL_zero(request);
+
+    request.freq = AUDIO_FREQUENCY;
+    request.format = AUDIO_S16;
+    request.channels = AUDIO_N_CHANNELS;
+    request.callback = audio_callback;
+    /* TODO: check for errors */
+    ctx.audio_device = SDL_OpenAudioDevice(NULL, 0, &request, &got, 0);
+    ctx.audio_stream_format = got.format;
+    ctx.audio_stream_frequency = got.freq;
+    ctx.audio_stream_channel_count = got.channels;
+
+    SDL_PauseAudioDevice(ctx.audio_device, 0);
+
+    }
+
     /* images */
     if (IMG_Init(IMG_INIT_PNG) == 0) {
         CRY_SDL("SDL_image initialization failed.");
diff --git a/src/private/audio.h b/src/private/audio.h
new file mode 100644
index 0000000..16a4344
--- /dev/null
+++ b/src/private/audio.h
@@ -0,0 +1,49 @@
+#ifndef PRIVATE_AUDIO_H
+#define PRIVATE_AUDIO_H
+
+#include "../audio.h"
+
+#include <SDL2/SDL_audio.h>
+
+#define STB_VORBIS_HEADER_ONLY
+#include <stb_vorbis.c>
+
+#include <stdbool.h>
+#include <stdint.h>
+
+
+typedef enum audio_file_type {
+    audio_file_type_ogg,
+    audio_file_type_xm,
+    audio_file_type_count,
+    audio_file_type_unknown,
+} t_audio_file_type;
+
+
+union audio_context {
+    struct {
+        stb_vorbis *handle;
+        unsigned char *data;
+        int frequency;
+        uint8_t channel_count;
+    } vorbis;
+};
+
+
+struct audio_channel {
+    t_play_audio_args args;
+    enum audio_file_type file_type;
+    union audio_context context; /* interpreted by `file_type` value */
+    const char *path;
+};
+
+
+struct audio_channel_pair {
+    char *key;
+    struct audio_channel value;
+};
+
+
+void audio_callback(void *userdata, uint8_t *stream, int len);
+
+#endif
diff --git a/src/util.c b/src/util.c
index 6509132..5551eee 100644
--- a/src/util.c
+++ b/src/util.c
@@ -101,7 +101,7 @@ void *ccalloc(size_t num, size_t size) {
 }
 
 
-int64_t file_to_bytes(char *path, unsigned char **buf_out) {
+int64_t file_to_bytes(const char *path, unsigned char **buf_out) {
     SDL_RWops *handle = PHYSFSRWOPS_openRead(path);
 
     if (handle == NULL) {
@@ -119,7 +119,7 @@ int64_t file_to_bytes(char *path, unsigned char **buf_out) {
 }
 
 
-char *file_to_str(char *path) {
+char *file_to_str(const char *path) {
     SDL_RWops *handle = PHYSFSRWOPS_openRead(path);
 
     if (handle == NULL) {
diff --git a/src/util.h b/src/util.h
index c6bd2e1..bbdb150 100644
--- a/src/util.h
+++ b/src/util.h
@@ -51,10 +51,10 @@ void *ccalloc(size_t num, size_t size);
 
 /* sets buf_out to a pointer to a byte buffer which must be freed. */
 /* returns the size of this buffer. */
-int64_t file_to_bytes(char *path, unsigned char **buf_out);
+int64_t file_to_bytes(const char *path, unsigned char **buf_out);
 
 /* returns a pointer to a string which must be freed */
-char *file_to_str(char *path);
+char *file_to_str(const char *path);
 
 
 /* returns true if str ends with suffix */