Feature/wav decoding support in oac_demo by akashtc · Pull Request #10 · AOMediaCodec/oac

akashtc · 2026-02-05T00:42:36Z

Implemented WAV file support for encoding in oac_demo.
Added a test WAV file to the tests directory.
Updated the
README
to reflect the new functionality.
Successfully built and tested the changes.

Keep QEXT just for 96 kHz input/output Signed-off-by: Jean-Marc Valin <jeanmarcv@google.com>

Remove ENABLE_RES24 option, keep QEXT just for 96 kHz support Signed-off-by: Jean-Marc Valin <jeanmarcv@google.com>

We now use 32-bit precision, but MIPS code matches older 16-bit implementation.

These will break every time we change the bitstream

Now always using ec_tell_frac() to check the number of remaining bits instead of attempting our own accounting, which is more complicated and could (in extremely rare cases) fail.

Also update condition for coding the trim symbol

Also update condition for coding the pitch symbols

Backport from Opus commit a3f0ec02 (gitlab MR 190)

jmvalin · 2026-02-05T20:25:12Z

src/oac_demo.c

+#include "oac.h"
+#include "oac_multistream.h"
+#include "oac_private.h"
+#include "oac_types.h"


Why do you need to change the order?

Still not addressed

src/oac_demo.c

jmvalin · 2026-02-05T20:26:29Z

src/oac_demo.c

+      int size;
+      int sample_size = 2;
+      if (format == FORMAT_S24_LE)
+        sample_size = 3;


As per coding style, have the statement on the same line as the if as in the original code

not addressed

jmvalin · 2026-02-05T20:27:03Z

src/oac_demo.c

-        fseek(fin, 0, SEEK_SET);
-        mode_switch_time = size/sample_size/channels/nb_modes_in_list;
-        fprintf(stderr, "Switching mode every %d samples\n", mode_switch_time);
+    const char *ext = strrchr(inFile, '.');


declaration after statement. We're still in C90

jmvalin · 2026-02-05T20:28:58Z

src/oac_demo.c

+                channels);
+        fprintf(stderr, "Using parameters from command line.\n");
+      }
+    } else if (mode_list) {


Don't understand why that needs to be in an else for the wav code. mode_list is a totally independent thing.

jmvalin · 2026-02-05T20:29:29Z

src/oac_demo.c

-        fprintf(stderr, "Switching mode every %d samples\n", mode_switch_time);
+    const char *ext = strrchr(inFile, '.');
+    if (!decode_only && ext && strcmp(ext, ".wav") == 0) {
+      char header[44];


Indenting is 4-space

still not addressed

jmvalin · 2026-02-05T20:33:21Z

src/oac_demo.c

+        goto failure;
+      }
+
+      oac_int32 wav_format_samplerate = *(oac_int32 *)(header + 24);


This kind of casting is not legal in C and it would break on a big endian machine. Use char_to_int() instead.

jmvalin · 2026-02-05T20:34:12Z

src/oac_demo.c

+      }
+
+      oac_int32 wav_format_samplerate = *(oac_int32 *)(header + 24);
+      short wav_format_channels = *(short *)(header + 22);


See above, but you'll need your own 16-bit conversion code.

jmvalin · 2026-02-05T20:35:26Z

src/oac_demo.c

+                wav_format_samplerate, wav_format_channels, sampling_rate,
+                channels);
+        fprintf(stderr, "Using parameters from command line.\n");
+      }


You'll also want to check and/or auto-detect the PCM format (16-bit, 24-bit, float).

jmvalin · 2026-02-05T20:36:21Z

README

  -loss <perc>      : simulate packet loss, in percent (0-100); default: 0

-input and output are little-endian signed 16-bit PCM files or OAC
+input and output are little-endian signed 16-bit PCM files, WAV files (for decoding), or OAC


Shouldn't it say "for encoding" since wav input is for encode?

jmvalin · 2026-02-05T20:38:48Z

One feature of wav files is they can be made of multiple chunks, which would break your current code. I don't think we should necessarily handle multiple chunks, but it would be good to not silently do something crazy when that happens. For example, we could just detect that and saw we don't support it, or encode just the first chunk, or something else that sane. Right now the next chunks may just end up interpreted as noise.

jmvalin · 2026-02-05T21:40:50Z

src/oac_demo.c

        fprintf (stderr, "Could not open input file %s\n", argv[argc - 2]);
        goto failure;
    }
+    const char *ext;


Still a declaration after statement.

jmvalin · 2026-02-05T21:41:54Z

src/oac_demo.c

+      unsigned char data_size_bytes[4];
+      oac_int32 riff_chunk_size;
+      oac_int32 data_chunk_size;
+      short bits_per_sample;


Best to use the oac_int16 type (or just "int" here would work) because short doesn't have a fixed length.

jmvalin · 2026-02-05T21:42:16Z

src/oac_demo.c


+// Helper function for little-endian 16-bit
+static short char_to_short(unsigned char ch[2]) {
+  return (short)(((unsigned short)ch[1] << 8) | (unsigned short)ch[0]);


please use oac_int16 instead of "short" here.

jmvalin · 2026-02-05T21:44:54Z

src/oac_demo.c

+        case 24:
+          format = FORMAT_S24_LE;
+          break;
+        case 32: // Assuming float for 32-bit


Please don't use // comments (C90 again)

jmvalin · 2026-02-05T21:45:41Z

src/oac_demo.c

-        fseek(fin, 0, SEEK_SET);
-        mode_switch_time = size/sample_size/channels/nb_modes_in_list;
-        fprintf(stderr, "Switching mode every %d samples\n", mode_switch_time);
+      int size;


Please don't reindent the code.

akashtc

Check now

debargha · 2026-02-12T18:28:00Z

So this patch only handles encode, correct?

akashtc · 2026-02-17T19:37:36Z

So this patch only handles encode, correct?

Yes

jmvalin · 2026-02-18T15:32:59Z

You might want to rebase your PR. You can do it on top of main or on top of oac-next.

jmvalin and others added 24 commits January 15, 2026 11:10

Removing quality extension layer

0082297

Keep QEXT just for 96 kHz input/output Signed-off-by: Jean-Marc Valin <jeanmarcv@google.com>

Always use extra precision from RES24 and QEXT

e4f10df

Remove ENABLE_RES24 option, keep QEXT just for 96 kHz support Signed-off-by: Jean-Marc Valin <jeanmarcv@google.com>

Rename identifiers from opus to oac

fd0c105

Rename downloads

e6c46a1

Remove reference to Opus logo

54e4d9f

Fix version script

7ec547f

Reset LT version

f9b6096

Reformat to OAC coding style

e7b0367

Remove pre-rfc8251 compatibility

520f8ce

README update

b65aabe

Fix gitlab references

7a9dc21

Add Windows rename script

51bdcf3

Remove outdated MIPS optimization

c581f50

We now use 32-bit precision, but MIPS code matches older 16-bit implementation.

Disable decoder fixed final range checks

5542712

These will break every time we change the bitstream

Cleaning up remaining bits calculations

c60d1a2

Now always using ec_tell_frac() to check the number of remaining bits instead of attempting our own accounting, which is more complicated and could (in extremely rare cases) fail.

Tune spreading CDF

ff2d47d

Tune the trim CDF

80408d6

Also update condition for coding the trim symbol

Tuning tapset CDF

8ee3a05

Also update condition for coding the pitch symbols

Fix sample_max computation

5468b2c

Backport from Opus commit a3f0ec02 (gitlab MR 190)

Renaming internal symbols to have an oaci_ prefix

d8c8e0c

Renaming static symbols to have an oaci_ prefix

ce3f7ed

Rename more static symbols that are often inlined

dc3d2f2

feat: Add WAV file decoding support to oac_demo

411f6e0

docs: Update README and add test WAV file

ad0fc34

akashtc changed the base branch from main to oac-next February 5, 2026 19:22

jmvalin reviewed Feb 5, 2026

View reviewed changes

src/oac_demo.c Show resolved Hide resolved

jmvalin reviewed Feb 5, 2026

View reviewed changes

Akash Choudhari added 3 commits February 5, 2026 13:21

Apply review comments for WAV decoding

8e35bd9

Resolve TODOs for WAV handling: format detection and multi-chunk warning

3f89398

Update README to reflect WAV format detection

b0d4916

jmvalin reviewed Feb 5, 2026

View reviewed changes

Fix remaining style and indentation issues from review

6b8f018

akashtc commented Feb 5, 2026

View reviewed changes

akashtc marked this pull request as draft February 5, 2026 23:53

akashtc self-assigned this Feb 5, 2026

jmvalin force-pushed the oac-next branch from dc3d2f2 to bc9317c Compare February 17, 2026 22:44

Conversation

akashtc commented Feb 5, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jmvalin commented Feb 5, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

akashtc left a comment

Choose a reason for hiding this comment

Uh oh!

debargha commented Feb 12, 2026

Uh oh!

akashtc commented Feb 17, 2026

Uh oh!

jmvalin commented Feb 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Comments