af_scaletempo2: fix audio artifact on initial WSOLA iteration

The first WSOLA iteration overlapped audio with whatever was in the
`wsola_output` buffer. This was either silence (if not run before), or
old frames (if switching to 1x and back to a different speed).

Track the state of the output buffer and memcpy the whole window for the
first iteration instead.

1 files changed, 3 insertions, 0 deletions

diff --git a/audio/filter/af_scaletempo2_internals.h b/audio/filter/af_scaletempo2_internals.h
index b062159966..64f4104019 100644
--- a/audio/filter/af_scaletempo2_internals.h
+++ b/audio/filter/af_scaletempo2_internals.h
@@ -80,6 +80,9 @@ struct mp_scaletempo2 {
     // them and can be copied to output if fill_buffer() is called. It also
     // specifies the index where the next WSOLA window has to overlap-and-add.
     int num_complete_frames;
+    // Whether |wsola_output| contains an additional |ola_hop_size| of overlap
+    // frames for the next iteration.
+    bool wsola_output_started;
     // Overlap-and-add window.
     float *ola_window;
     // Transition window, used to update |optimal_block| by a weighted sum of