zyc 689fa8936b Integrate Volcengine realtime voice + Live2D mouth driving
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-14 15:39:23 +08:00

456 lines
18 KiB
Objective-C

//
// VadViewController.m
// SpeechDemo
//
// Created by bytedance on 2023/1/30.
// Copyright © 2023 tianlei.richard. All rights reserved.
//
#import "VadViewController.h"
#import <AVFoundation/AVFoundation.h>
#import "AppDelegate.h"
#import "FileUtils.h"
#import "SettingsHelper.h"
#import "ViewController.h"
#import "SensitiveDefines.h"
@interface VadViewController () <SpeechEngineDelegate, UITextViewDelegate>
@property (weak, nonatomic) IBOutlet UITextView *resultTextView;
@property (weak, nonatomic) IBOutlet UITextField *statusTextView;
@property (weak, nonatomic) IBOutlet UIButton *engineInitButton;
@property (weak, nonatomic) IBOutlet UIButton *engineUninitButton;
@property (weak, nonatomic) IBOutlet UIButton *startEngineButton;
@property (weak, nonatomic) IBOutlet UIButton *stopEngineButton;
@property (weak, nonatomic) IBOutlet UIButton *longPressButton;
@property (strong, nonatomic) SpeechEngine *curEngine;
@property (assign, nonatomic) BOOL engineStarted;
@property (nonatomic, strong) NSString *deviceID;
@property (nonatomic, assign) long talkingFinisheTimestamp;
@property (nonatomic, assign) long vadDuration;
@property (nonatomic, assign) long vadBeginPosition;
@property (nonatomic, assign) long vadEndPosition;
@property (strong, nonatomic) NSString *debugPath;
@property (weak, nonatomic) StreamRecorder *streamRecorder;
// settings
@property (strong, nonatomic) Settings *settings;
@end
@implementation VadViewController
- (void)viewDidLoad {
[super viewDidLoad];
self.settings = [[SettingsHelper shareInstance]getSettings:VIEW_VAD];
self.engineInitButton.enabled = TRUE;
self.engineUninitButton.enabled = FALSE;
self.startEngineButton.enabled = FALSE;
self.stopEngineButton.enabled = FALSE;
self.longPressButton.enabled = FALSE;
[self.statusTextView setText:@"Waiting for init."];
[self decorateTextView:self.resultTextView];
[ViewController setAppDelegate:(AppDelegate *)[[UIApplication sharedApplication] delegate]];
UILongPressGestureRecognizer *longPgr = [[UILongPressGestureRecognizer alloc] initWithTarget:self
action:@selector(longPressTriggered:)];
longPgr.minimumPressDuration = 0.5;
[self.longPressButton addGestureRecognizer:longPgr];
self.streamRecorder = [ViewController getStreamRecorder];
self.engineStarted = FALSE;
}
- (void)viewDidDisappear:(BOOL)animated {
[self uninitEngine];
[super viewDidDisappear:animated];
}
- (void)decorateTextView:(UITextView *)textView {
textView.layer.cornerRadius = 5.0f;
textView.layer.borderWidth = .25f;
textView.layer.borderColor = [UIColor grayColor].CGColor;
}
#pragma mark - SpeechEngineDelegate
- (void)onMessageWithType:(SEMessageType)type andData:(NSData *)data {
NSLog(@"Message Type: %d.", type);
switch (type) {
case SEEngineStart:
[self speechEngineStarted];
break;
case SEEngineStop:
[self speechEngineStopped];
break;
case SEEngineError:
[self speechEngineError:data];
break;
case SEVadSilence:
break;
case SEVadSil2Speech:
[self messageVadBegin:data];
break;
case SEVadSpeech:
[self messasgeVadSpeech:data];
break;
case SEVadSpeech2Sil:
[self messageVadEnd:data];
break;
case SEVadAudioData:
self.vadDuration += data.length;
break;
case SEEngineLog:
NSLog(@"engine log: %s", (char*)data.bytes);
break;
default:
break;
}
}
#pragma mark - UI Actions
- (IBAction)initEngine:(id)sender {
[self initEngine];
}
- (IBAction)uninitEngine:(id)sender {
if (self.engineStarted) {
[self.statusTextView setText:@"Engine is busy, stop it first!"];
return;
}
[self uninitEngine];
[self.resultTextView setTextColor:UIColor.grayColor];
[self.resultTextView setText:@"点击或按住说话后,展示音频活性检测结果"];
}
- (IBAction)startEngine:(id)sender {
NSLog(@"Start engine.");
NSString* appID = [self.settings getString:SETTING_APPID];
[self.curEngine setStringParam:appID.length <= 0 ? SDEF_APPID : appID forKey:SE_PARAMS_KEY_APP_ID_STRING];
[self.curEngine setBoolParam:[self.settings getBool:SETTING_RESTART_AUDIO_SESSION_ENABLE]
forKey:SE_PARAMS_KEY_RESTART_AUDIOSESSION_BOOL];
[self.curEngine setBoolParam:[self.settings getBool:SETTING_RESUME_OTHERS_INTERRUPTED_PLAYBACK_ENABLE]
forKey:SE_PARAMS_KEY_RESUME_OTHERS_INTERRUPTED_PLAYBACK_BOOL];
[self.curEngine setBoolParam:[self.settings getBool:SETTING_GET_VOLUME] forKey:SE_PARAMS_KEY_ENABLE_GET_VOLUME_BOOL];
[self.curEngine setIntParam:[self.settings getInt:SETTING_VAD_HEAD_SILENCE_THRESHOLD] forKey:SE_PARAMS_KEY_VAD_HEAD_SILENCE_THRESHOLD_INT];
[self.curEngine setIntParam:[self.settings getInt:SETTING_VAD_TAIL_SILENCE_THRESHOLD] forKey:SE_PARAMS_KEY_VAD_TAIL_SILENCE_THRESHOLD_INT];
if ([[self getRecorderType] isEqualToString:SE_RECORDER_TYPE_STREAM]) {
[self.curEngine setIntParam:[self.streamRecorder getSampleRate] forKey:SE_PARAMS_KEY_CUSTOM_SAMPLE_RATE_INT];
if (![self.streamRecorder start]) {
[self speechEngineNoPermission];
return;
}
} else if ([[self getRecorderType] isEqualToString:SE_RECORDER_TYPE_FILE]) {
NSString* file_path = [NSString stringWithFormat:@"%@/%@", self.debugPath, @"vad_rec_file.pcm"];
NSLog(@"test file path: %@", file_path);
[self.curEngine setStringParam:file_path forKey:SE_PARAMS_KEY_RECORDER_FILE_STRING];
}
SEEngineErrorCode ret = [self.curEngine sendDirective:SEDirectiveStartEngine];
if (ret == SERecCheckEnvironmentFailed) {
[self speechEngineNoPermission];
}
}
- (IBAction)stopEngine:(id)sender {
NSLog(@"Stop engine.");
[self.curEngine sendDirective:SEDirectiveStopEngine];
}
- (void)longPressTriggered:(UILongPressGestureRecognizer *)longPgr {
if (longPgr.state == UIGestureRecognizerStateBegan) {
NSLog(@"Long press begin.");
[self setResultText:@""];
NSString* appID = [self.settings getString:SETTING_APPID];
[self.curEngine setStringParam:appID.length <= 0 ? SDEF_APPID : appID forKey:SE_PARAMS_KEY_APP_ID_STRING];
[self.curEngine setBoolParam:[self.settings getBool:SETTING_RESTART_AUDIO_SESSION_ENABLE]
forKey:SE_PARAMS_KEY_RESTART_AUDIOSESSION_BOOL];
[self.curEngine setBoolParam:[self.settings getBool:SETTING_RESUME_OTHERS_INTERRUPTED_PLAYBACK_ENABLE]
forKey:SE_PARAMS_KEY_RESUME_OTHERS_INTERRUPTED_PLAYBACK_BOOL];
[self.curEngine setBoolParam:[self.settings getBool:SETTING_GET_VOLUME]
forKey:SE_PARAMS_KEY_ENABLE_GET_VOLUME_BOOL];
[self.curEngine setIntParam:[self.settings getInt:SETTING_VAD_HEAD_SILENCE_THRESHOLD] forKey:SE_PARAMS_KEY_VAD_HEAD_SILENCE_THRESHOLD_INT];
[self.curEngine setIntParam:[self.settings getInt:SETTING_VAD_TAIL_SILENCE_THRESHOLD] forKey:SE_PARAMS_KEY_VAD_TAIL_SILENCE_THRESHOLD_INT];
if ([[self getRecorderType] isEqualToString:SE_RECORDER_TYPE_STREAM]) {
[self.curEngine setIntParam:[self.streamRecorder getSampleRate] forKey:SE_PARAMS_KEY_CUSTOM_SAMPLE_RATE_INT];
if (![self.streamRecorder start]) {
[self speechEngineNoPermission];
return;
}
} else if ([[self getRecorderType] isEqualToString:SE_RECORDER_TYPE_FILE]) {
NSString* file_path = [NSString stringWithFormat:@"%@/%@", self.debugPath, @"vad_rec_file.pcm"];
NSLog(@"test file path: %@", file_path);
[self.curEngine setStringParam:file_path forKey:SE_PARAMS_KEY_RECORDER_FILE_STRING];
}
SEEngineErrorCode ret = [self.curEngine sendDirective:SEDirectiveStartEngine];
if (ret == SERecCheckEnvironmentFailed) {
[self speechEngineNoPermission];
}
} else if (longPgr.state == UIGestureRecognizerStateEnded) {
NSLog(@"Long press ended.");
self.talkingFinisheTimestamp = [[NSDate date] timeIntervalSince1970] * 1000;
[self.curEngine sendDirective:SEDirectiveFinishTalking];
if ([[self getRecorderType] isEqualToString:SE_RECORDER_TYPE_STREAM]) {
[self.streamRecorder stop];
}
}
}
#pragma mark - Init Methods
- (void)initEngine {
AppDelegate *appDelegate = [ViewController getAppDelegate];
if (appDelegate == nil) {
appDelegate = (AppDelegate *)[[UIApplication sharedApplication] delegate];
}
if (appDelegate.deviceID.length < 1) {
self.engineInitButton.enabled = FALSE;
dispatch_async(dispatch_get_main_queue(), ^{
[self.statusTextView setText:@"Waiting for get deviceID."];
sleep(1);
[self initEngine];
});
return;
}
[ViewController setAppDelegate:appDelegate];
self.deviceID = appDelegate.deviceID;
if (self.curEngine == nil) {
self.curEngine = [[SpeechEngine alloc] init];
}
if (![self.curEngine createEngineWithDelegate:self]) {
NSLog(@"Create speech engine failed.");
return;
}
[self.resultTextView setTextColor:UIColor.blackColor];
NSLog(@"Engine version: %@", [self.curEngine getVersion]);
[self initOfflineModel:^() {
self.debugPath = NSSearchPathForDirectoriesInDomains(NSDocumentDirectory, NSUserDomainMask, YES).firstObject;
NSLog(@"Debug path: %@", self.debugPath);
[self.curEngine setStringParam:self.debugPath forKey:SE_PARAMS_KEY_DEBUG_PATH_STRING];
[self.curEngine setStringParam:SE_LOG_LEVEL_TRACE forKey:SE_PARAMS_KEY_LOG_LEVEL_STRING];
[self.curEngine setStringParam:SDEF_APPID forKey:SE_PARAMS_KEY_APP_ID_STRING];
[self.curEngine setStringParam:self.deviceID forKey:SE_PARAMS_KEY_DEVICE_ID_STRING];
[self.curEngine setIntParam:1 forKey:SE_PARAMS_KEY_CHANNEL_NUM_INT];
[self.curEngine setBoolParam:[self.settings getBool:SETTING_GET_VOLUME]
forKey:SE_PARAMS_KEY_ENABLE_GET_VOLUME_BOOL];
[self.curEngine setStringParam:@"" forKey:SE_PARAMS_KEY_VAD_REC_PATH_STRING];
if ([self.settings getBool:SETTING_VAD_RECORDER_SAVE]) {
[self.curEngine setStringParam:self.debugPath forKey:SE_PARAMS_KEY_VAD_REC_PATH_STRING];
}
[self.curEngine setIntParam:4000 forKey:SE_PARAMS_KEY_VAD_HEAD_SILENCE_THRESHOLD_INT];
[self.curEngine setIntParam:2000 forKey:SE_PARAMS_KEY_VAD_TAIL_SILENCE_THRESHOLD_INT];
[self.curEngine setStringParam:[self getRecorderType] forKey:SE_PARAMS_KEY_RECORDER_TYPE_STRING];
[self.curEngine setStringParam:SE_VAD_ENGINE forKey:SE_PARAMS_KEY_ENGINE_NAME_STRING];
[self.curEngine setIntParam:16000 forKey:SE_PARAMS_KEY_SAMPLE_RATE_INT];
NSString* aedResourcePath = [[SpeechResourceManager shareInstance] getModelPath:SE_AED_MODEL];
NSLog(@"petrel aed resource path: %@", aedResourcePath);
[self.curEngine setStringParam:aedResourcePath forKey:SE_PARAMS_KEY_AED_RESOURCE_PATH_STRING];
if ([[self getRecorderType] isEqualToString:SE_RECORDER_TYPE_STREAM]) {
if ([self.streamRecorder getSampleRate] != 16000) {
[self.curEngine setBoolParam:TRUE forKey:SE_PARAMS_KEY_ENABLE_RESAMPLER_BOOL];
}
}
SEEngineErrorCode ret = [self.curEngine initEngine];
if (ret != SENoError) {
NSLog(@"Init Engine failed: %d", ret);
}
if (ret == SENoError) {
[self speechEngineInitOk];
} else {
[self speechEngineInitFailed:ret];
}
} fail:^{
[self speechEngineInitFailed:kSERDownloadFailed];
}];
}
- (NSString *)getRecorderType {
SettingOptions* recorderTypeOptions = [self.settings getOptions:SETTING_RECORD_TYPE];
switch (recorderTypeOptions.chooseIdx) {
case 0:
return SE_RECORDER_TYPE_RECORDER;
case 1:
return SE_RECORDER_TYPE_FILE;
case 2:
return SE_RECORDER_TYPE_STREAM;
default:
break;
}
return @"";
}
- (void)initOfflineModel:(void(^)(void))succ fail:(void(^)(void))fail {
NSString *model = SE_AED_MODEL;
SpeechResourceManager *speechResourceManager = [SpeechResourceManager shareInstance];
[speechResourceManager checkModelVersion:model completion:^(SEResourceStatus status, BOOL needUpdate, NSData *data) {
NSLog(@"Need update: %@", needUpdate ? @"YES" : @"NO");
if (status != kSERSuccess || needUpdate == NO) {
if ([speechResourceManager checkModelExist:model]) {
NSLog(@"Model exist!");
succ();
} else {
NSLog(@"Model not exist!");
fail();
}
return;
}
// need to update model
[speechResourceManager fetchModelByName:model completion:^(SEResourceStatus status, NSData* data) {
NSLog(@"Completion: %@", status == kSERSuccess ? @"success" : @"fail");
if (status == kSERSuccess) {
succ();
} else {
NSLog(@"Failed: %d", status);
fail();
}
}];
}];
}
- (void)uninitEngine {
[self.curEngine destroyEngine];
self.curEngine = nil;
self.engineInitButton.enabled = TRUE;
self.engineUninitButton.enabled = FALSE;
self.startEngineButton.enabled = FALSE;
self.stopEngineButton.enabled = FALSE;
self.longPressButton.enabled = FALSE;
}
#pragma mark - Engine Callback
- (void)speechEngineNoPermission {
dispatch_async(dispatch_get_main_queue(), ^{
[self uninitEngine];
[self.statusTextView setText:@"No permission!"];
self.engineInitButton.enabled = TRUE;
self.engineUninitButton.enabled = FALSE;
});
}
- (void)speechEngineInitOk {
[self.streamRecorder setSpeechEngine:self.curEngine];
dispatch_async(dispatch_get_main_queue(), ^{
[self.statusTextView setText:@"Ready"];
[self.resultTextView setText:[NSString stringWithFormat:@"DeviceID: %@", self.deviceID]];
self.engineUninitButton.enabled = TRUE;
self.engineInitButton.enabled = FALSE;
self.startEngineButton.enabled = TRUE;
self.longPressButton.enabled = TRUE;
});
}
- (void)speechEngineInitFailed:(int)initStatus {
dispatch_async(dispatch_get_main_queue(), ^{
[self uninitEngine];
[self.statusTextView setText:[[NSString alloc] initWithFormat:@"Failed to init engine, %d!", initStatus]];
self.engineInitButton.enabled = TRUE;
self.engineUninitButton.enabled = FALSE;
});
}
- (void)speechEngineStarted {
dispatch_async(dispatch_get_main_queue(), ^{
self.engineStarted = true;
self.vadDuration = 0;
self.vadBeginPosition = 0;
self.vadEndPosition = 0;
[self.statusTextView setText:@"Engine Started!"];
self.startEngineButton.enabled = FALSE;
self.stopEngineButton.enabled = TRUE;
self.longPressButton.enabled = FALSE;
});
}
- (void)speechEngineStopped {
dispatch_async(dispatch_get_main_queue(), ^{
if ([[self getRecorderType] isEqualToString:SE_RECORDER_TYPE_STREAM]) {
[self.streamRecorder stop];
}
self.engineStarted = FALSE;
[self.statusTextView setText:@"Engine Stopped!"];
self.startEngineButton.enabled = TRUE;
self.stopEngineButton.enabled = FALSE;
self.longPressButton.enabled = TRUE;
});
}
- (void)speechEngineError:(NSData *)data {
dispatch_async(dispatch_get_main_queue(), ^{
[self.resultTextView setText:[[NSString alloc] initWithData:data encoding:NSUTF8StringEncoding]];
[self stopEngine:nil];
});
}
- (void)messageVadBegin:(NSData *)data {
id json_obj = [NSJSONSerialization JSONObjectWithData:data options:NSJSONReadingMutableContainers error:nil];
if ([json_obj isKindOfClass:[NSDictionary class]]) {
NSDictionary *vad_info = json_obj;
self.vadBeginPosition = [[vad_info objectForKey:@"start"] doubleValue];
}
dispatch_async(dispatch_get_main_queue(), ^{
[self.resultTextView setText:[NSString stringWithFormat:@"Vad begin, bos: %ld", self.vadBeginPosition]];
});
}
- (void)messasgeVadSpeech:(NSData *)data {
id json_obj = [NSJSONSerialization JSONObjectWithData:data options:NSJSONReadingMutableContainers error:nil];
if ([json_obj isKindOfClass:[NSDictionary class]]) {
NSDictionary *vad_info = json_obj;
double currentSegEnd = [[vad_info objectForKey:@"end"] doubleValue];
if (currentSegEnd > self.vadEndPosition) {
self.vadEndPosition = currentSegEnd;
}
}
}
- (void)messageVadEnd:(NSData *)data {
dispatch_async(dispatch_get_main_queue(), ^{
[self.resultTextView setText:[NSString stringWithFormat:@"Vad end, eos: %ld, speech duration: %ldms.", self.vadEndPosition, self.vadDuration/(2*16000/1000)]];
[self stopEngine:nil];
});
}
- (void)setResultText:(NSString *)result {
dispatch_async(dispatch_get_main_queue(), ^{
[self.resultTextView setText:[result stringByTrimmingCharactersInSet:[NSCharacterSet whitespaceCharacterSet]]];
});
}
#pragma mark - UITextViewDelegate
- (BOOL)textView:(UITextView *)textView shouldChangeTextInRange:(NSRange)range replacementText:(NSString *)text {
if([text isEqualToString:@"\n"]) {
[textView resignFirstResponder];
return NO;
}
return YES;
}
#pragma mark - Navigation
// In a storyboard-based application, you will often want to do a little preparation before navigation
- (void)prepareForSegue:(UIStoryboardSegue *)segue sender:(id)sender {
// Get the new view controller using [segue destinationViewController].
// Pass the selected object to the new view controller.
id nextPage = [segue destinationViewController];
[nextPage setValue:VIEW_VAD forKey:@"viewId"];
}
@end