Files
talk/internal/audio/mp3_test.go
T

123 lines
8.1 KiB
Go
Raw Normal View History

package audio
import (
"encoding/base64"
"strings"
"testing"
)
// Two real MP3 fixtures generated once (ffmpeg on the dev host, embedded in
// the test so tests stay free of tool dependencies at runtime). Both match
// the delivery format of the cloud providers:
//
// fixtureMP3Mono: 0.15 s silence, 24 kHz mono 48 kbit/s; exactly the
// Edge TTS output format (audio-24khz-48kbitrate-mono-
// mp3); exercises the decoder's mono-to-stereo path.
// fixtureMP3Stereo: 0.15 s 440 Hz sine, 44.1 kHz stereo 64 kbit/s; the
// typical ElevenLabs/OpenAI format.
const fixtureMP3MonoB64 = "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"
const fixtureMP3StereoB64 = "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"
func fixture(t *testing.T, b64 string) []byte {
t.Helper()
data, err := base64.StdEncoding.DecodeString(b64)
if err != nil {
t.Fatalf("decoding fixture: %v", err)
}
return data
}
// TestDecodeMP3Mono: the Edge format (24 kHz mono) decodes to 16-bit PCM
// stereo; the decoder mirrors mono onto two channels, samples are aligned
// to 4 bytes (2 channels x 2 bytes).
func TestDecodeMP3Mono(t *testing.T) {
f, pcm, err := decodeMP3(fixture(t, fixtureMP3MonoB64))
if err != nil {
t.Fatalf("decodeMP3 (mono): %v", err)
}
if f.SampleRate != 24000 {
t.Errorf("SampleRate = %d, want 24000", f.SampleRate)
}
if f.Channels != 2 || f.BitsPerCoded != 16 || f.Float {
t.Errorf("format = %+v, want 2 channels/16 bit/int", f)
}
if len(pcm) == 0 || len(pcm)%4 != 0 {
t.Errorf("PCM length = %d, want > 0 and divisible by 4 (16-bit stereo samples)", len(pcm))
}
}
// TestDecodeMP3Stereo: the ElevenLabs/OpenAI-typical format (44.1 kHz stereo).
func TestDecodeMP3Stereo(t *testing.T) {
f, pcm, err := decodeMP3(fixture(t, fixtureMP3StereoB64))
if err != nil {
t.Fatalf("decodeMP3 (stereo): %v", err)
}
if f.SampleRate != 44100 || f.Channels != 2 || f.BitsPerCoded != 16 {
t.Errorf("format = %+v, want 44100 Hz stereo 16 bit", f)
}
if len(pcm) == 0 || len(pcm)%4 != 0 {
t.Errorf("PCM length = %d, want > 0 and divisible by 4", len(pcm))
}
}
// TestDecodeMP3Errors: data that looks like MP3 but contains no frames is a
// clear mp3 error (not the WAV error path).
func TestDecodeMP3Errors(t *testing.T) {
cases := []struct {
name string
data []byte
want string
}{
{"garbage with frame sync", []byte{0xFF, 0xFB, 'n', 'o', 't', ' ', 'm', 'p', '3'}, "mp3:"},
{"ID3 tag without frames", []byte("ID3\x04\x00\x00\x00\x00\x00\x00"), "mp3:"},
}
for _, tc := range cases {
t.Run(tc.name, func(t *testing.T) {
_, _, err := decodeMP3(tc.data)
if err == nil || !strings.Contains(err.Error(), tc.want) {
t.Errorf("decodeMP3 error = %v, want message containing %q", err, tc.want)
}
})
}
}
// TestLooksLikeMP3 pins the format detection.
func TestLooksLikeMP3(t *testing.T) {
yes := [][]byte{
fixture(t, fixtureMP3StereoB64), // ID3 tag
{0xFF, 0xFB, 0x00, 0x00}, // raw frame sync
}
no := [][]byte{
[]byte("RIFF....WAVE"),
[]byte("not audio"),
{0xFF},
{},
}
for _, d := range yes {
if !looksLikeMP3(d) {
t.Errorf("looksLikeMP3(%x) = false, want true", d[:min(4, len(d))])
}
}
for _, d := range no {
if looksLikeMP3(d) {
t.Errorf("looksLikeMP3(%q) = true, want false", d)
}
}
}
// TestDecodeAudioSwitch: RIFF goes to the WAV parser, MP3 to the MP3
// decoder, everything else falls through with the WAV error.
func TestDecodeAudioSwitch(t *testing.T) {
f, _, err := decodeAudio(buildWAV(1, 1, 22050, 16, []byte{1, 2, 3, 4}))
if err != nil || f.SampleRate != 22050 {
t.Errorf("decodeAudio(WAV) = %+v, %v; should take the WAV path", f, err)
}
f, _, err = decodeAudio(fixture(t, fixtureMP3MonoB64))
if err != nil || f.SampleRate != 24000 {
t.Errorf("decodeAudio(MP3) = %+v, %v; should take the MP3 path", f, err)
}
if _, _, err = decodeAudio([]byte("not audio")); err == nil || !strings.Contains(err.Error(), "wav:") {
t.Errorf("decodeAudio(garbage) = %v, want WAV error", err)
}
}