static void Main(string[] args)
{
using (IDeepSpeech sttClient = new DeepSpeech(DEEPSPEECH_MODEL))
{
sttClient.EnableExternalScorer(DEEPSPEECH_SCORER);
string AudioFilePath = @"audio\2830-3980-0043.wav";
WaveBuffer waveBuffer = new WaveBuffer(File.ReadAllBytes(AudioFilePath));
string speechResult = sttClient.SpeechToText(
waveBuffer.ShortBuffer,
Convert.ToUInt32(waveBuffer.MaxSize / 2));
Console.WriteLine(speechResult);
waveBuffer.Clear();
}
}
static void Main(string[] args)
{
string model = null;
string lm = null;
string trie = null;
string audio = null;
bool extended = false;
if (args.Length > 0)
{
model = GetArgument(args, "--model");
lm = GetArgument(args, "--lm");
trie = GetArgument(args, "--trie");
audio = GetArgument(args, "--audio");
extended = !string.IsNullOrWhiteSpace(GetArgument(args, "--extended"));
}
const uint BEAM_WIDTH = 500;
const float LM_ALPHA = 0.75f;
const float LM_BETA = 1.85f;
Stopwatch stopwatch = new Stopwatch();
using (IDeepSpeech sttClient = new DeepSpeech())
{
try
{
Console.WriteLine("Loading model...");
stopwatch.Start();
sttClient.CreateModel(
model ?? "output_graph.pbmm",
BEAM_WIDTH);
stopwatch.Stop();
Console.WriteLine($"Model loaded - {stopwatch.Elapsed.Milliseconds} ms");
stopwatch.Reset();
if (lm != null)
{
Console.WriteLine("Loadin LM...");
sttClient.EnableDecoderWithLM(
lm ?? "lm.binary",
trie ?? "trie",
LM_ALPHA, LM_BETA);
}
string audioFile = audio ?? "arctic_a0024.wav";
var waveBuffer = new WaveBuffer(File.ReadAllBytes(audioFile));
using (var waveInfo = new WaveFileReader(audioFile))
{
Console.WriteLine("Running inference....");
stopwatch.Start();
string speechResult;
if (extended)
{
Metadata metaResult = sttClient.SpeechToTextWithMetadata(waveBuffer.ShortBuffer, Convert.ToUInt32(waveBuffer.MaxSize / 2));
speechResult = MetadataToString(metaResult);
}
else
{
speechResult = sttClient.SpeechToText(waveBuffer.ShortBuffer, Convert.ToUInt32(waveBuffer.MaxSize / 2));
}
stopwatch.Stop();
Console.WriteLine($"Audio duration: {waveInfo.TotalTime.ToString()}");
Console.WriteLine($"Inference took: {stopwatch.Elapsed.ToString()}");
Console.WriteLine((extended ? $"Extended result: " : "Recognized text: ") + speechResult);
}
waveBuffer.Clear();
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
}
}
}
static void Main(string[] args)
{
string model = null;
string alphabet = null;
string lm = null;
string trie = null;
string audio = null;
if (args.Length > 0)
{
model = GetArgument(args, "--model");
alphabet = GetArgument(args, "--alphabet");
lm = GetArgument(args, "--lm");
trie = GetArgument(args, "--trie");
audio = GetArgument(args, "--audio");
}
const uint N_CEP = 26;
const uint N_CONTEXT = 9;
const uint BEAM_WIDTH = 200;
const float LM_ALPHA = 0.75f;
const float LM_BETA = 1.85f;
Stopwatch stopwatch = new Stopwatch();
using (IDeepSpeech sttClient = new DeepSpeech())
{
var result = 1;
Console.WriteLine("Loading model...");
stopwatch.Start();
try
{
result = sttClient.CreateModel(
model ?? "output_graph.pbmm",
N_CEP, N_CONTEXT,
alphabet ?? "alphabet.txt",
BEAM_WIDTH);
}
catch (IOException ex)
{
Console.WriteLine("Error loading lm.");
Console.WriteLine(ex.Message);
}
stopwatch.Stop();
if (result == 0)
{
Console.WriteLine($"Model loaded - {stopwatch.Elapsed.Milliseconds} ms");
stopwatch.Reset();
if (lm != null)
{
Console.WriteLine("Loadin LM...");
try
{
result = sttClient.EnableDecoderWithLM(
alphabet ?? "alphabet.txt",
lm ?? "lm.binary",
trie ?? "trie",
LM_ALPHA, LM_BETA);
}
catch (IOException ex)
{
Console.WriteLine("Error loading lm.");
Console.WriteLine(ex.Message);
}
}
string audioFile = audio ?? "arctic_a0024.wav";
var waveBuffer = new WaveBuffer(File.ReadAllBytes(audioFile));
using (var waveInfo = new WaveFileReader(audioFile))
{
Console.WriteLine("Running inference....");
stopwatch.Start();
string speechResult = sttClient.SpeechToText(waveBuffer.ShortBuffer, Convert.ToUInt32(waveBuffer.MaxSize / 2), 16000);
stopwatch.Stop();
Console.WriteLine($"Audio duration: {waveInfo.TotalTime.ToString()}");
Console.WriteLine($"Inference took: {stopwatch.Elapsed.ToString()}");
Console.WriteLine($"Recognized text: {speechResult}");
}
waveBuffer.Clear();
}
else
{
Console.WriteLine("Error loding the model.");
}
}
}
static void Main(string[] args)
{
string model = null;
string scorer = null;
string audio = null;
bool extended = false;
if (args.Length > 0)
{
model = GetArgument(args, "--model");
scorer = GetArgument(args, "--scorer");
audio = GetArgument(args, "--audio");
extended = !string.IsNullOrWhiteSpace(GetArgument(args, "--extended"));
}
Stopwatch stopwatch = new Stopwatch();
try
{
Console.WriteLine("Loading model...");
stopwatch.Start();
// sphinx-doc: csharp_ref_model_start
using IDeepSpeech sttClient = new DeepSpeech(model ?? "output_graph.pbmm");
// sphinx-doc: csharp_ref_model_stop
stopwatch.Stop();
Console.WriteLine($"Model loaded - {stopwatch.Elapsed.Milliseconds} ms");
stopwatch.Reset();
if (scorer != null)
{
Console.WriteLine("Loading scorer...");
sttClient.EnableExternalScorer(scorer ?? "kenlm.scorer");
}
string audioFile = audio ?? "new-home-in-the-stars-16k.wav";
var waveBuffer = new WaveBuffer(File.ReadAllBytes(audioFile));
using (var waveInfo = new WaveFileReader(audioFile))
{
Console.WriteLine("Running inference....");
stopwatch.Start();
string speechResult;
// sphinx-doc: csharp_ref_inference_start
if (extended)
{
Metadata metaResult = sttClient.SpeechToTextWithMetadata(waveBuffer.ShortBuffer,
Convert.ToUInt32(waveBuffer.MaxSize / 2), 1);
speechResult = MetadataToString(metaResult.Transcripts[0]);
}
else
{
speechResult = sttClient.SpeechToText(waveBuffer.ShortBuffer,
Convert.ToUInt32(waveBuffer.MaxSize / 2));
}
// sphinx-doc: csharp_ref_inference_stop
stopwatch.Stop();
Console.WriteLine($"Audio duration: {waveInfo.TotalTime.ToString()}");
Console.WriteLine($"Inference took: {stopwatch.Elapsed.ToString()}");
Console.WriteLine((extended ? $"Extended result: " : "Recognized text: ") + speechResult);
}
waveBuffer.Clear();
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
}
}
public static string Main1(string s)
{
string model = null;
string audio = null;
string hotwords = null;
bool extended = false;
model = "C:/Users/Nour El-Din/Documents/deepspeech/deepspeech-0.9.3-models.pbmm";
var scorer = "C:/Users/Nour El-Din/Documents/deepspeech/deepspeech-0.9.3-models.scorer";
audio = "C:/Users/Nour El-Din/Documents/deepspeech/audio";
//hotwords = GetArgument(args, "--hot_words");
extended = true;
Stopwatch stopwatch = new Stopwatch();
try
{
Console.WriteLine("Loading model...");
stopwatch.Start();
// sphinx-doc: csharp_ref_model_start
using (IDeepSpeech sttClient = new DeepSpeech("C:/Users/Nour El-Din/Documents/deepspeech/deepspeech-0.9.3-models.pbmm"))
{
// sphinx-doc: csharp_ref_model_stop
stopwatch.Stop();
Console.WriteLine($"Model loaded - {stopwatch.Elapsed.Milliseconds} ms");
stopwatch.Reset();
if (scorer != null)
{
Console.WriteLine("Loading scorer...");
sttClient.EnableExternalScorer("C:/Users/Nour El-Din/Documents/deepspeech/deepspeech-0.9.3-models.scorer");
}
if (hotwords != null)
{
Console.WriteLine($"Adding hot-words {hotwords}");
char[] sep = { ',' };
string[] word_boosts = hotwords.Split(sep);
foreach (string word_boost in word_boosts)
{
char[] sep1 = { ':' };
string[] word = word_boost.Split(sep1);
sttClient.AddHotWord(word[0], float.Parse(word[1]));
}
}
Directory.SetCurrentDirectory("C:/Users/Nour El-Din/Downloads/Graduation-Project-Dragons-master/Graduation Project Dragons/wwwroot/Video");
string v_Name = s.Split('.')[0];
var enviroment = System.Environment.CurrentDirectory;
string projectDirectory = Directory.GetParent(enviroment).Parent.FullName;
Console.WriteLine(projectDirectory);
string m_Path = @$ "C:/Users/Nour El-Din/Downloads/Graduation-Project-Dragons-master/Graduation Project Dragons/wwwroot/Video/{v_Name}.mp3";
bool mp3_Found = File.Exists(m_Path) ? true : false;
string w_Path = @$ "C:/Users/Nour El-Din/Downloads/Graduation-Project-Dragons-master/Graduation Project Dragons/wwwroot/Video/{v_Name}.wav";
bool wav_Found = File.Exists(w_Path) ? true : false;
string strCmdText;
if (!mp3_Found)
{
strCmdText = $"/c ffmpeg -i {v_Name}.mp4 {v_Name}.mp3";
System.Diagnostics.Process.Start("CMD.exe", strCmdText);
}
else if (!wav_Found)
{
strCmdText = $"/c ffmpeg -i {v_Name}.mp3 -acodec pcm_s16le -ac 1 -ar 16000 {v_Name}.wav";
System.Diagnostics.Process.Start("CMD.exe", strCmdText);
}
else
{
string audioFile = $"C:/Users/Nour El-Din/Downloads/Graduation-Project-Dragons-master/Graduation Project Dragons/wwwroot/Video/{v_Name}.wav";
var waveBuffer = new WaveBuffer(File.ReadAllBytes(audioFile));
using (var waveInfo = new WaveFileReader(audioFile))
{
Console.WriteLine("Running inference....");
stopwatch.Start();
string w = "";
string speechResult;
// sphinx-doc: csharp_ref_inference_start
if (extended)
{
Metadata metaResult = sttClient.SpeechToTextWithMetadata(waveBuffer.ShortBuffer,
Convert.ToUInt32(waveBuffer.MaxSize / 2), 1);
speechResult = MetadataToString(metaResult.Transcripts[0]);
data new_Data = new data();
List <data> full_Text = new List <data>();
double word_Start = 0;;
double word_End;
string temp = "";
int counter = 0;
int counter2 = 0;
foreach (var c in metaResult.Transcripts[0].Tokens)
{
counter2++;
if (counter == 0)
{
word_Start = c.StartTime;
counter++;
}
temp += c.Text;
if (c.Text == " ")
{
counter = 0;
word_End = c.StartTime;
new_Data.start = word_Start;
new_Data.end = word_End;
new_Data.word = temp;
full_Text.Add(new_Data);
temp = "";
new_Data = new data();
continue;
}
w += c.Text;
if (counter2 == metaResult.Transcripts[0].Tokens.Length - 1)
{
word_End = c.StartTime;
new_Data.start = word_Start;
new_Data.end = word_End;
new_Data.word = temp;
full_Text.Add(new_Data);
temp = "";
}
}
}
else
{
speechResult = sttClient.SpeechToText(waveBuffer.ShortBuffer,
Convert.ToUInt32(waveBuffer.MaxSize / 2));
}
// sphinx-doc: csharp_ref_inference_stop
stopwatch.Stop();
Console.WriteLine($"Audio duration: {waveInfo.TotalTime.ToString()}");
Console.WriteLine($"Inference took: {stopwatch.Elapsed.ToString()}");
Console.WriteLine((extended ? $"Extended result: " : "Recognized text: ") + speechResult);
return(w);
}
waveBuffer.Clear();
}
}
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
}
return(" ");
}
public static string Main1(string s)
{
string model = null;
bool extended = false;
model = "C:/Users/Nour El-Din/Documents/deepspeech/deepspeech-0.9.3-models.pbmm";
var scorer = "C:/Users/Nour El-Din/Documents/deepspeech/deepspeech-0.9.3-models.scorer";
extended = true;
string wwwPath = HomeController.Environment.WebRootPath;
string path = Path.Combine(HomeController.Environment.WebRootPath, "Video");
Stopwatch stopwatch = new Stopwatch();
try
{
Console.WriteLine("Loading model...");
stopwatch.Start();
// sphinx-doc: csharp_ref_model_start
using (IDeepSpeech sttClient = new DeepSpeech(model))
{
// sphinx-doc: csharp_ref_model_stop
stopwatch.Stop();
Console.WriteLine($"Model loaded - {stopwatch.Elapsed.Milliseconds} ms");
stopwatch.Reset();
if (scorer != null)
{
Console.WriteLine("Loading scorer...");
sttClient.EnableExternalScorer(scorer);
}
Directory.SetCurrentDirectory(path);
string v_Name = s.Split('.')[0];
string x = path + @$ "/{v_Name}.mp4";
string m_Path = path + @$ "/{v_Name}.mp3";
bool mp3_Found = File.Exists(m_Path) ? true : false;
string w_Path = path + @$ "/{v_Name}.wav";
bool wav_Found = File.Exists(w_Path) ? true : false;
string strCmdText;
if (!mp3_Found)
{
strCmdText = $"/c ffmpeg -i {v_Name}.mp4 {v_Name}.mp3";
Process process = new Process();
process.StartInfo.FileName = "cmd.exe";
process.StartInfo.WindowStyle = ProcessWindowStyle.Hidden;
process.StartInfo.Arguments = strCmdText;
process.Start();
process.WaitForExit();
}
if (!wav_Found)
{
strCmdText = $"/c ffmpeg -i {v_Name}.mp3 -acodec pcm_s16le -ac 1 -ar 16000 {v_Name}.wav";
Process process = new Process();
process.StartInfo.FileName = "cmd.exe";
process.StartInfo.WindowStyle = ProcessWindowStyle.Hidden;
process.StartInfo.Arguments = strCmdText;
process.Start();
process.WaitForExit();
}
string audioFile = path + $"/{v_Name}.wav";
var waveBuffer = new WaveBuffer(File.ReadAllBytes(audioFile));
using (var waveInfo = new WaveFileReader(audioFile))
{
Console.WriteLine("Running inference....");
stopwatch.Start();
string w = "";
string speechResult;
// sphinx-doc: csharp_ref_inference_start
if (extended)
{
Metadata metaResult = sttClient.SpeechToTextWithMetadata(waveBuffer.ShortBuffer,
Convert.ToUInt32(waveBuffer.MaxSize / 2), 1);
speechResult = MetadataToString(metaResult.Transcripts[0]);
data new_Data = new data();
List <data> full_Text = new List <data>();
double word_Start = 0;;
double word_End;
string temp = "";
int counter = 0;
int counter2 = 0;
int count = 0;
foreach (var c in metaResult.Transcripts[0].Tokens)
{
counter2++;
if (counter == 0)
{
word_Start = c.StartTime;
counter++;
}
temp += c.Text;
if (c.Text == " ")
{
counter = 0;
word_End = c.StartTime;
new_Data.start = word_Start;
new_Data.end = word_End;
new_Data.word = temp;
full_Text.Add(new_Data);
w += temp + ',';
w += (double)word_Start + ",";
w += (double)word_End + "/";
temp = "";
new_Data = new data();
continue;
}
//w += c.Text;
if (counter2 == metaResult.Transcripts[0].Tokens.Length)
{
word_End = c.StartTime;
new_Data.start = word_Start;
new_Data.end = word_End;
new_Data.word = temp;
full_Text.Add(new_Data);
w += temp + ',';
w += (double)word_Start + ",";
w += (double)word_End + "/";
temp = "";
}
count++;
}
}
else
{
speechResult = sttClient.SpeechToText(waveBuffer.ShortBuffer,
Convert.ToUInt32(waveBuffer.MaxSize / 2));
}
// sphinx-doc: csharp_ref_inference_stop
stopwatch.Stop();
Console.WriteLine($"Audio duration: {waveInfo.TotalTime.ToString()}");
Console.WriteLine($"Inference took: {stopwatch.Elapsed.ToString()}");
Console.WriteLine((extended ? $"Extended result: " : "Recognized text: ") + speechResult);
waveBuffer.Clear();
return(w);
}
}
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
}
return(" ");
}