This commit is contained in:
intelligence committed 2026-08-29 23:11:50 +02:00
1 parent 32e7f32e0a
commit 523acd3349
23 files changed
+482 -98

No files matched your search

+12
View File
@@ -14,6 +14,12 @@ android {
versionName '0.1.1'
testInstrumentationRunner 'android.test.InstrumentationTestRunner'
externalNativeBuild {
cmake {
arguments '-DANDROID_STL=none'
}
}
}
signingConfigs {
@@ -46,6 +52,12 @@ android {
checkReleaseBuilds true
warningsAsErrors true
}
externalNativeBuild {
cmake {
path file('src/main/cpp/CMakeLists.txt')
}
}
}
dependencies {
+24
View File
@@ -0,0 +1,24 @@
cmake_minimum_required(VERSION 3.22.1)
project(transui_mp3 LANGUAGES C)
set(SHINE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/../../../../../third_party/shine/src/lib")
set(TRANSUI_MP3_DIR "${CMAKE_CURRENT_SOURCE_DIR}/../../../../../third_party/transui_mp3")
add_library(transui_mp3 SHARED
native_mp3_encoder.c
"${TRANSUI_MP3_DIR}/transui_mp3.c"
"${SHINE_DIR}/bitstream.c"
"${SHINE_DIR}/huffman.c"
"${SHINE_DIR}/l3bitstream.c"
"${SHINE_DIR}/l3loop.c"
"${SHINE_DIR}/l3mdct.c"
"${SHINE_DIR}/l3subband.c"
"${SHINE_DIR}/layer3.c"
"${SHINE_DIR}/reservoir.c"
"${SHINE_DIR}/tables.c"
)
target_compile_features(transui_mp3 PRIVATE c_std_11)
target_compile_options(transui_mp3 PRIVATE -Wall -Wextra)
target_include_directories(transui_mp3 PRIVATE "${TRANSUI_MP3_DIR}" "${SHINE_DIR}")
target_link_libraries(transui_mp3 PRIVATE m)
@@ -0,0 +1,33 @@
#include <jni.h>
#include "transui_mp3.h"
JNIEXPORT jbyteArray JNICALL
Java_io_github_transui_android_Mp3Encoder_encodeNative(JNIEnv *env,
jclass type,
jbyteArray pcm) {
(void)type;
if (!pcm) return NULL;
const jsize pcm_size = (*env)->GetArrayLength(env, pcm);
jbyte *pcm_bytes = (*env)->GetByteArrayElements(env, pcm, NULL);
if (!pcm_bytes) return NULL;
uint8_t *mp3 = NULL;
size_t mp3_size = 0;
const int result = transui_mp3_encode_mono_16k_32kbps(
(const uint8_t *)pcm_bytes, (size_t)pcm_size, &mp3, &mp3_size);
(*env)->ReleaseByteArrayElements(env, pcm, pcm_bytes, JNI_ABORT);
if (result != TRANSUI_MP3_OK || !mp3 || mp3_size > INT32_MAX) {
transui_mp3_free(mp3);
return NULL;
}
jbyteArray output = (*env)->NewByteArray(env, (jsize)mp3_size);
if (output) {
(*env)->SetByteArrayRegion(env, output, 0, (jsize)mp3_size,
(const jbyte *)mp3);
}
transui_mp3_free(mp3);
return output;
}
@@ -0,0 +1,22 @@
package io.github.transui.android;
final class Mp3Encoder {
static {
System.loadLibrary("transui_mp3");
}
private Mp3Encoder() {}
static byte[] encodePcm16Mono(byte[] pcm) {
if (pcm == null || pcm.length == 0 || pcm.length % 2 != 0) {
throw new IllegalArgumentException("PCM input must contain complete 16-bit samples");
}
byte[] mp3 = encodeNative(pcm);
if (mp3 == null || mp3.length == 0) {
throw new IllegalStateException("MP3 encoding failed");
}
return mp3;
}
private static native byte[] encodeNative(byte[] pcm);
}
@@ -19,11 +19,11 @@ final class TranscriptionClient {
private static final int MAX_RESPONSE_BYTES = 1024 * 1024;
private volatile HttpURLConnection activeConnection;
Result transcribe(AppConfig config, byte[] wav) {
Result transcribe(AppConfig config, byte[] mp3) {
if (config.endpoint.isEmpty()) {
return Result.error("No transcription endpoint is configured", SpeechRecognizer.ERROR_CLIENT);
}
return request(config.endpoint, config.model, config.apiKey, config.prompt, wav);
return request(config.endpoint, config.model, config.apiKey, config.prompt, mp3);
}
void cancel() {
@@ -36,11 +36,11 @@ final class TranscriptionClient {
String model,
String key,
String prompt,
byte[] wav) {
byte[] mp3) {
HttpURLConnection connection = null;
try {
String boundary = "transui-" + UUID.randomUUID();
byte[] requestBody = multipartBody(boundary, model, prompt, wav);
byte[] requestBody = multipartBody(boundary, model, prompt, mp3);
connection = (HttpURLConnection) new URL(endpointUrl(endpoint)).openConnection();
activeConnection = connection;
connection.setConnectTimeout(12_000);
@@ -86,17 +86,17 @@ final class TranscriptionClient {
return result + "/v1/audio/transcriptions";
}
private static byte[] multipartBody(String boundary, String model, String prompt, byte[] wav)
private static byte[] multipartBody(String boundary, String model, String prompt, byte[] mp3)
throws IOException {
ByteArrayOutputStream bytes = new ByteArrayOutputStream(wav.length + 1024);
ByteArrayOutputStream bytes = new ByteArrayOutputStream(mp3.length + 1024);
try (DataOutputStream output = new DataOutputStream(bytes)) {
writeField(output, boundary, "response_format", "json");
if (!model.isEmpty()) writeField(output, boundary, "model", model);
if (!prompt.isEmpty()) writeField(output, boundary, "prompt", prompt);
output.writeBytes("--" + boundary + "\r\n");
output.writeBytes("Content-Disposition: form-data; name=\"file\"; filename=\"recording.wav\"\r\n");
output.writeBytes("Content-Type: audio/wav\r\n\r\n");
output.write(wav);
output.writeBytes("Content-Disposition: form-data; name=\"file\"; filename=\"recording.mp3\"\r\n");
output.writeBytes("Content-Type: audio/mpeg\r\n\r\n");
output.write(mp3);
output.writeBytes("\r\n--" + boundary + "--\r\n");
}
return bytes.toByteArray();
@@ -124,8 +124,15 @@ public final class TransuiRecognitionService extends RecognitionService {
if (pcm == null || cancelled.get()) return;
sendEndOfSpeech();
if (cancelled.get()) return;
TranscriptionClient.Result result = client.transcribe(
config, WavEncoder.encodePcm16Mono(pcm, SAMPLE_RATE));
byte[] mp3;
try {
mp3 = Mp3Encoder.encodePcm16Mono(pcm);
} catch (IllegalArgumentException | IllegalStateException error) {
Log.e(TAG, "MP3 encoding failed", error);
sendError(callback, SpeechRecognizer.ERROR_AUDIO);
return;
}
TranscriptionClient.Result result = client.transcribe(config, mp3);
if (cancelled.get()) return;
if (result.text == null) {
Log.w(TAG, result.error);
@@ -67,8 +67,14 @@ final class VoiceRecordingController {
byte[] pcm = captureAudio();
if (pcm == null || cancelled.get()) return;
listener.onTranscribing();
TranscriptionClient.Result result = client.transcribe(
config, WavEncoder.encodePcm16Mono(pcm, SAMPLE_RATE));
byte[] mp3;
try {
mp3 = Mp3Encoder.encodePcm16Mono(pcm);
} catch (IllegalArgumentException | IllegalStateException error) {
listener.onError(SpeechRecognizer.ERROR_AUDIO);
return;
}
TranscriptionClient.Result result = client.transcribe(config, mp3);
if (cancelled.get()) return;
if (result.text != null) {
listener.onResult(result.text);
@@ -1,44 +0,0 @@
package io.github.transui.android;
import java.io.ByteArrayOutputStream;
final class WavEncoder {
private static final int HEADER_SIZE = 44;
private WavEncoder() {}
static byte[] encodePcm16Mono(byte[] pcm, int sampleRate) {
ByteArrayOutputStream output = new ByteArrayOutputStream(HEADER_SIZE + pcm.length);
writeAscii(output, "RIFF");
writeInt(output, 36 + pcm.length);
writeAscii(output, "WAVE");
writeAscii(output, "fmt ");
writeInt(output, 16);
writeShort(output, 1);
writeShort(output, 1);
writeInt(output, sampleRate);
writeInt(output, sampleRate * 2);
writeShort(output, 2);
writeShort(output, 16);
writeAscii(output, "data");
writeInt(output, pcm.length);
output.write(pcm, 0, pcm.length);
return output.toByteArray();
}
private static void writeAscii(ByteArrayOutputStream output, String value) {
for (int index = 0; index < value.length(); index++) output.write(value.charAt(index));
}
private static void writeShort(ByteArrayOutputStream output, int value) {
output.write(value & 0xff);
output.write((value >>> 8) & 0xff);
}
private static void writeInt(ByteArrayOutputStream output, int value) {
output.write(value & 0xff);
output.write((value >>> 8) & 0xff);
output.write((value >>> 16) & 0xff);
output.write((value >>> 24) & 0xff);
}
}
@@ -1,35 +0,0 @@
package io.github.transui.android;
import static org.junit.Assert.assertArrayEquals;
import static org.junit.Assert.assertEquals;
import java.nio.charset.StandardCharsets;
import org.junit.Test;
public final class WavEncoderTest {
@Test
public void writesPcm16MonoHeaderAndPayload() {
byte[] pcm = new byte[]{1, 2, 3, 4};
byte[] wav = WavEncoder.encodePcm16Mono(pcm, 16_000);
assertEquals("RIFF", ascii(wav, 0, 4));
assertEquals(40, littleEndianInt(wav, 4));
assertEquals("WAVE", ascii(wav, 8, 4));
assertEquals(16_000, littleEndianInt(wav, 24));
assertEquals(32_000, littleEndianInt(wav, 28));
assertEquals(4, littleEndianInt(wav, 40));
assertArrayEquals(pcm, new byte[]{wav[44], wav[45], wav[46], wav[47]});
}
private static String ascii(byte[] bytes, int offset, int length) {
return new String(bytes, offset, length, StandardCharsets.US_ASCII);
}
private static int littleEndianInt(byte[] bytes, int offset) {
return (bytes[offset] & 0xff)
| ((bytes[offset + 1] & 0xff) << 8)
| ((bytes[offset + 2] & 0xff) << 16)
| ((bytes[offset + 3] & 0xff) << 24);
}
}