diff --git a/apps/codec_thread.c b/apps/codec_thread.c
index 137ceffdc0..40c064e86b 100644
--- a/apps/codec_thread.c
+++ b/apps/codec_thread.c
@@ -249,7 +249,7 @@ static void codec_pcmbuf_insert_callback(
         }
         else
         {
-            dsp_process(ci.dsp, &src, &dst);
+            dsp_process(ci.dsp, &src, &dst, true);
 
             if (dst.remcount > 0)
             {
diff --git a/apps/plugin.h b/apps/plugin.h
index a0c41a20e7..0e56257b2c 100644
--- a/apps/plugin.h
+++ b/apps/plugin.h
@@ -751,7 +751,7 @@ struct plugin_api {
                               unsigned int setting, intptr_t value);
     struct dsp_config * (*dsp_get_config)(unsigned int dsp_id);
     void (*dsp_process)(struct dsp_config *dsp, struct dsp_buffer *src,
-                        struct dsp_buffer *dst);
+                        struct dsp_buffer *dst, bool thread_yield);
 
     enum channel_status (*mixer_channel_status)(enum pcm_mixer_channel channel);
     const void * (*mixer_channel_get_buffer)(enum pcm_mixer_channel channel,
diff --git a/apps/plugins/midi/midiplay.c b/apps/plugins/midi/midiplay.c
index 04636e7e5b..72fda4f572 100644
--- a/apps/plugins/midi/midiplay.c
+++ b/apps/plugins/midi/midiplay.c
@@ -444,7 +444,7 @@ static inline void synthbuf(void)
         dst.remcount = 0;
         dst.bufcount = available;
         dst.p16out = (int16_t *)outptr;
-        rb->dsp_process(dsp, &src, &dst);
+        rb->dsp_process(dsp, &src, &dst, true);
         if (dst.remcount > 0)
         {
             outptr += dst.remcount;
diff --git a/apps/plugins/mpegplayer/audio_thread.c b/apps/plugins/mpegplayer/audio_thread.c
index d1fdcfd4f3..be57d46e26 100644
--- a/apps/plugins/mpegplayer/audio_thread.c
+++ b/apps/plugins/mpegplayer/audio_thread.c
@@ -665,7 +665,7 @@ static void audio_thread(void)
             }
 
             dst.bufcount = size / (2 * sizeof (int16_t));
-            rb->dsp_process(td.dsp, &td.src, &dst);
+            rb->dsp_process(td.dsp, &td.src, &dst, true);
 
             if (dst.remcount > 0)
             {
diff --git a/apps/plugins/test_codec.c b/apps/plugins/test_codec.c
index e33a04c44d..95dcb687cb 100644
--- a/apps/plugins/test_codec.c
+++ b/apps/plugins/test_codec.c
@@ -230,7 +230,7 @@ static int process_dsp(const void *ch1, const void *ch2, int count)
     while (1)
     {
         int old_remcount = dst.remcount;
-        rb->dsp_process(ci.dsp, &src, &dst);
+        rb->dsp_process(ci.dsp, &src, &dst, true);
         
         if (dst.bufcount <= 0 ||
             (src.remcount <= 0 && dst.remcount <= old_remcount))
diff --git a/apps/rbcodecconfig.h b/apps/rbcodecconfig.h
index 9777ded345..a8a78133d1 100644
--- a/apps/rbcodecconfig.h
+++ b/apps/rbcodecconfig.h
@@ -33,11 +33,14 @@ struct dsp_loop_context
 #endif
 };
 
-static inline void dsp_process_start(struct dsp_loop_context *ctx)
+static inline void dsp_process_start(struct dsp_loop_context *ctx, bool thread_yield)
 {
     /* At least perform one yield before starting */
     ctx->last_yield = current_tick;
-    yield();
+    if (thread_yield)
+    {
+        yield();
+    }
 #if defined(CPU_COLDFIRE)
     /* set emac unit for dsp processing, and save old macsr, we're running in
        codec thread context at this point, so can't clobber it */
@@ -46,14 +49,17 @@ static inline void dsp_process_start(struct dsp_loop_context *ctx)
 #endif
 }
 
-static inline void dsp_process_loop(struct dsp_loop_context *ctx)
+static inline void dsp_process_loop(struct dsp_loop_context *ctx, bool thread_yield)
 {
     /* Yield at least once each tick */
     long tick = current_tick;
     if (TIME_AFTER(tick, ctx->last_yield))
     {
         ctx->last_yield = tick;
-        yield();
+        if (thread_yield)
+        {
+            yield();
+        }
     }
 }
 
@@ -66,12 +72,12 @@ static inline void dsp_process_end(struct dsp_loop_context *ctx)
     (void)ctx;
 }
 
-#define DSP_PROCESS_START() \
+#define DSP_PROCESS_START(yield) \
     struct dsp_loop_context __ctx; \
-    dsp_process_start(&__ctx)
+    dsp_process_start(&__ctx, yield)
 
-#define DSP_PROCESS_LOOP() \
-    dsp_process_loop(&__ctx)
+#define DSP_PROCESS_LOOP(yield) \
+    dsp_process_loop(&__ctx, yield)
 
 #define DSP_PROCESS_END() \
     dsp_process_end(&__ctx)
diff --git a/apps/voice_thread.c b/apps/voice_thread.c
index dfc2dd8ed6..0f0d89c24e 100644
--- a/apps/voice_thread.c
+++ b/apps/voice_thread.c
@@ -511,7 +511,7 @@ static enum voice_state voice_buffer_insert(struct voice_thread_data *td)
         dst.bufcount = VOICE_PCM_FRAME_COUNT;
 
         td->dst = &dst;
-        dsp_process(td->dsp, &td->src, &dst);
+        dsp_process(td->dsp, &td->src, &dst, true);
         td->dst = NULL;
 
         voice_buf_commit(dst.remcount);
diff --git a/docs/PLUGIN_API b/docs/PLUGIN_API
index 961d3da2b8..330d165982 100644
--- a/docs/PLUGIN_API
+++ b/docs/PLUGIN_API
@@ -664,11 +664,12 @@ int32_t dsp_get_timestretch(void)
     \return
     \description
 
-void dsp_process(struct dsp_config *dsp, struct dsp_buffer *src, struct dsp_buffer *dst)
+void dsp_process(struct dsp_config *dsp, struct dsp_buffer *src, struct dsp_buffer *dst, bool thread_yield)
     \group sound
     \param dsp
     \param src
     \param dst
+    \param thread_yield
     \description
 
 void dsp_set_crossfeed_type(int type)
diff --git a/firmware/usbstack/usb_audio.c b/firmware/usbstack/usb_audio.c
index 239aea4d7c..321f4405c1 100644
--- a/firmware/usbstack/usb_audio.c
+++ b/firmware/usbstack/usb_audio.c
@@ -42,6 +42,7 @@
 #include "settings.h"
 #include "core_alloc.h"
 #include "pcm_mixer.h"
+#include "dsp_core.h"
 
 #define LOGF_ENABLE
 #include "logf.h"
@@ -352,7 +353,7 @@ int tmp_saved_vol;
 static unsigned char *rx_buffer;
 int rx_buffer_handle;
 /* buffer size */
-static int rx_buf_size[NR_BUFFERS];
+static int rx_buf_size[NR_BUFFERS]; // only used for debug screen counter now
 /* index of the next buffer to play */
 static int rx_play_idx;
 /* index of the next buffer to fill */
@@ -362,6 +363,14 @@ bool playback_audio_underflow;
 /* usb overflow ? */
 bool usb_rx_overflow;
 
+/* dsp processing buffers */
+#define DSP_BUF_SIZE (BUFFER_SIZE*4) // arbitrarily x4
+#define REAL_DSP_BUF_SIZE   ALIGN_UP(DSP_BUF_SIZE, 32)
+static uint16_t *dsp_buf;
+int dsp_buf_handle;
+static int dsp_buf_size[NR_BUFFERS];
+struct dsp_config *dsp = NULL;
+
 /* feedback variables */
 #define USB_FRAME_MAX 0x7FF
 #define NR_SAMPLES_HISTORY 32
@@ -502,7 +511,7 @@ int usb_audio_request_buf(void)
     audio_stop();
 
     // attempt to allocate the receive buffers
-    rx_buffer_handle = core_alloc(NR_BUFFERS * REAL_BUF_SIZE);
+    rx_buffer_handle = core_alloc(REAL_BUF_SIZE);
     if (rx_buffer_handle < 0)
     {
         alloc_failed = true;
@@ -518,6 +527,23 @@ int usb_audio_request_buf(void)
         // get the pointer to the actual buffer location
         rx_buffer = core_get_data(rx_buffer_handle);
     }
+
+    dsp_buf_handle = core_alloc(NR_BUFFERS * REAL_DSP_BUF_SIZE);
+    if (dsp_buf_handle < 0)
+    {
+        alloc_failed = true;
+        rx_buffer_handle = core_free(rx_buffer_handle);
+        rx_buffer = NULL;
+        return -1;
+    }
+    else
+    {
+        alloc_failed = false;
+
+        core_pin(dsp_buf_handle);
+
+        dsp_buf = core_get_data(dsp_buf_handle);
+    }
     // logf("usbaudio: got buffer");
     return 0;
 }
@@ -527,6 +553,9 @@ void usb_audio_free_buf(void)
     // logf("usbaudio: free buffer");
     rx_buffer_handle = core_free(rx_buffer_handle);
     rx_buffer = NULL;
+
+    dsp_buf_handle = core_free(dsp_buf_handle);
+    dsp_buf = NULL;
 }
 
 int usb_audio_request_endpoints(struct usb_class_driver *drv)
@@ -644,10 +673,11 @@ static void playback_audio_get_more(const void **start, size_t *size)
         *size = 0;
         return;
     }
+
     /* give buffer and advance */
     logf("usbaudio: buf adv");
-    *start = rx_buffer + (rx_play_idx * REAL_BUF_SIZE);
-    *size = rx_buf_size[rx_play_idx];
+    *start = dsp_buf + (rx_play_idx * REAL_DSP_BUF_SIZE/sizeof(*dsp_buf));
+    *size = dsp_buf_size[rx_play_idx];
     rx_play_idx = (rx_play_idx + 1) % NR_BUFFERS;
 
     /* if usb RX buffers had overflowed, we can start to receive again
@@ -658,7 +688,7 @@ static void playback_audio_get_more(const void **start, size_t *size)
     {
         logf("usbaudio: recover usb rx overflow");
         usb_rx_overflow = false;
-        usb_drv_recv_nonblocking(out_iso_ep_adr, rx_buffer + (rx_usb_idx * REAL_BUF_SIZE), BUFFER_SIZE);
+        usb_drv_recv_nonblocking(out_iso_ep_adr, rx_buffer, BUFFER_SIZE);
     }
     restore_irq(oldlevel);
 }
@@ -697,7 +727,8 @@ static void usb_audio_start_playback(void)
     mixer_set_frequency(hw_freq_sampr[as_playback_freq_idx]);
     pcm_apply_settings();
     mixer_channel_set_amplitude(PCM_MIXER_CHAN_USBAUDIO, MIX_AMP_UNITY);
-    usb_drv_recv_nonblocking(out_iso_ep_adr, rx_buffer + (rx_usb_idx * REAL_BUF_SIZE), BUFFER_SIZE);
+
+    usb_drv_recv_nonblocking(out_iso_ep_adr, rx_buffer, BUFFER_SIZE);
 }
 
 static void usb_audio_stop_playback(void)
@@ -1150,6 +1181,15 @@ void usb_audio_init_connection(void)
 {
     logf("usbaudio: init connection");
 
+    dsp = dsp_get_config(CODEC_IDX_AUDIO);
+    dsp_configure(dsp, DSP_RESET, 0);
+    dsp_configure(dsp, DSP_SET_STEREO_MODE, STEREO_INTERLEAVED);
+    dsp_configure(dsp, DSP_SET_SAMPLE_DEPTH, 16);
+#ifdef HAVE_PITCHCONTROL
+    sound_set_pitch(PITCH_SPEED_100);
+    dsp_set_timestretch(PITCH_SPEED_100);
+#endif
+
     usb_as_playback_intf_alt = 0;
     set_playback_sampling_frequency(HW_SAMPR_DEFAULT);
     tmp_saved_vol = sound_current(SOUND_VOLUME);
@@ -1261,20 +1301,36 @@ bool usb_audio_fast_transfer_complete(int ep, int dir, int status, int length)
         logf("usbaudio: frame: %d bytes: %d", usb_drv_get_frame_number(), length);
         if(status != 0)
             return true; /* FIXME how to handle error here ? */
+
         /* store length, queue buffer */
         rx_buf_size[rx_usb_idx] = length;
-        rx_usb_idx = (rx_usb_idx + 1) % NR_BUFFERS;
 
         // debug screen counter
         samples_received = samples_received + length;
-    
+
+        // process through DSP right away!
+        struct dsp_buffer src;
+        src.remcount = length/4; // in samples
+        src.pin[0] = rx_buffer;
+        src.proc_mask = 0;
+
+        struct dsp_buffer dst;
+        dst.remcount = 0;
+        dst.bufcount = DSP_BUF_SIZE/4; // in samples
+        dst.p16out = dsp_buf + (rx_usb_idx * REAL_DSP_BUF_SIZE/sizeof(*dsp_buf)); // array index
+
+        dsp_process(dsp, &src, &dst, false);
+        dsp_buf_size[rx_usb_idx] = dst.remcount * 2 * sizeof(*dsp_buf); // need value in bytes
+
+        rx_usb_idx = (rx_usb_idx + 1) % NR_BUFFERS;
+
         /* guard against IRQ to avoid race with completion audio completion */
         int oldlevel = disable_irq_save();
         /* setup a new transaction except if we ran out of buffers */
         if(rx_usb_idx != rx_play_idx)
         {
             logf("usbaudio: new transaction");
-            usb_drv_recv_nonblocking(out_iso_ep_adr, rx_buffer + (rx_usb_idx*REAL_BUF_SIZE), BUFFER_SIZE);
+            usb_drv_recv_nonblocking(out_iso_ep_adr, rx_buffer, BUFFER_SIZE);
         }
         else
         {
diff --git a/lib/rbcodec/dsp/dsp_core.c b/lib/rbcodec/dsp/dsp_core.c
index 5108c9ef47..5b515e165a 100644
--- a/lib/rbcodec/dsp/dsp_core.c
+++ b/lib/rbcodec/dsp/dsp_core.c
@@ -37,8 +37,8 @@
 
 #ifndef DSP_PROCESS_START
 /* These do nothing if not previously defined */
-#define DSP_PROCESS_START()
-#define DSP_PROCESS_LOOP()
+#define DSP_PROCESS_START(yield)
+#define DSP_PROCESS_LOOP(yield)
 #define DSP_PROCESS_END()
 #endif /* !DSP_PROCESS_START */
 
@@ -433,7 +433,7 @@ static FORCE_INLINE void dsp_proc_call(struct dsp_proc_slot *s,
  * of the call and how they function internally.
  */
 void dsp_process(struct dsp_config *dsp, struct dsp_buffer *src,
-                 struct dsp_buffer *dst)
+                 struct dsp_buffer *dst, bool thread_yield)
 {
     if (dst->bufcount <= 0)
     {
@@ -441,7 +441,7 @@ void dsp_process(struct dsp_config *dsp, struct dsp_buffer *src,
         return;
     }
 
-    DSP_PROCESS_START();
+    DSP_PROCESS_START(thread_yield);
 
     /* Tag input with codec-specified sample format */
     src->format = dsp->io_data.format;
@@ -479,7 +479,7 @@ void dsp_process(struct dsp_config *dsp, struct dsp_buffer *src,
         dsp_advance_buffer32(buf, outcount);
         dsp_advance_buffer_output(dst, outcount);
 
-        DSP_PROCESS_LOOP();
+        DSP_PROCESS_LOOP(thread_yield);
     } /* while */
 
     DSP_PROCESS_END();
diff --git a/lib/rbcodec/dsp/dsp_core.h b/lib/rbcodec/dsp/dsp_core.h
index 9b09d981cc..6f78733551 100644
--- a/lib/rbcodec/dsp/dsp_core.h
+++ b/lib/rbcodec/dsp/dsp_core.h
@@ -136,7 +136,7 @@ unsigned int dsp_get_id(const struct dsp_config *dsp);
 
 /* Process the given buffer - see implementation in dsp.c for more */
 void dsp_process(struct dsp_config *dsp, struct dsp_buffer *src,
-                 struct dsp_buffer *dst);
+                 struct dsp_buffer *dst, bool thread_yield);
 
 /* Change DSP settings */
 intptr_t dsp_configure(struct dsp_config *dsp, unsigned int setting,
diff --git a/lib/rbcodec/test/warble.c b/lib/rbcodec/test/warble.c
index b931339d6c..a25deec5b8 100644
--- a/lib/rbcodec/test/warble.c
+++ b/lib/rbcodec/test/warble.c
@@ -455,7 +455,7 @@ static void ci_pcmbuf_insert(const void *ch1, const void *ch2, int count)
             dst.p16out = buf;
             dst.bufcount = out_count;
 
-            dsp_process(ci.dsp, &src, &dst);
+            dsp_process(ci.dsp, &src, &dst, true);
 
             if (dst.remcount > 0) {
                 if (mode == MODE_WRITE)