Merge pull request #4 from Azure/audiotranscription

AudioTranscription file structure
Azure · Jan 11, 2024 · 873998a · 873998a
2 parents 9645b36 + 5b044b3
commit 873998a
Show file tree

Hide file tree

Showing 16 changed files with 325 additions and 14 deletions.
diff --git a/.gitignore b/.gitignore
@@ -407,3 +407,4 @@ FodyWeavers.xsd
 
 # JetBrains Rider
 *.sln.iml
+src/AIHub/appsettings.Development.json
diff --git a/src/AIHub/Controllers/AudioTranscriptionController.cs b/src/AIHub/Controllers/AudioTranscriptionController.cs
@@ -0,0 +1,210 @@
+namespace MVCWeb.Controllers;
+using System;
+using System.Net.Http;
+using System.Net.Http.Headers;
+using System.Threading.Tasks;
+using Newtonsoft.Json;
+using Microsoft.AspNetCore.Mvc;
+using System.Net;
+using Newtonsoft.Json.Linq;
+using Microsoft.VisualBasic;
+
+public class AudioTranscriptionController : Controller
+{
+    private readonly ILogger<HomeController> _logger;
+    private readonly IConfiguration _config;
+    private string SpeechRegion;
+    private string SpeechSubscriptionKey;
+    private string storageconnstring;
+    private readonly BlobServiceClient blobServiceClient;
+    private readonly BlobContainerClient containerClient;
+    private readonly IEnumerable<BlobItem> blobs;
+    private Uri sasUri;
+
+
+    //Results
+    string result_message_front;
+
+
+
+    private AudioTranscriptionModel model;
+
+
+    public AudioTranscriptionController(IConfiguration config)
+    {
+        _config = config;
+        SpeechRegion = _config.GetValue<string>("AudioTranscription:SpeechLocation");
+        SpeechSubscriptionKey = _config.GetValue<string>("AudioTranscription:SpeechSubscriptionKey");
+        storageconnstring = _config.GetValue<string>("Storage:ConnectionString");
+        BlobServiceClient blobServiceClient = new BlobServiceClient(storageconnstring);
+        containerClient = blobServiceClient.GetBlobContainerClient(_config.GetValue<string>("AudioTranscription:ContainerName"));
+        sasUri = containerClient.GenerateSasUri(Azure.Storage.Sas.BlobContainerSasPermissions.Read, DateTimeOffset.UtcNow.AddHours(1));
+        // Obtiene una lista de blobs en el contenedor
+        blobs = containerClient.GetBlobs();
+        model = new AudioTranscriptionModel();
+    }
+
+    public IActionResult AudioTranscription()
+    {
+        return View();
+    }
+
+    [HttpPost]
+    public async Task<IActionResult> TranscribeAudio(string audio_url, IFormFile imageFile)
+    {
+
+        string audio = audio_url + sasUri.Query;
+
+    // CALL 1: STT 3.1
+
+        var client = new HttpClient();
+        var request = new HttpRequestMessage(HttpMethod.Post, "https://"+SpeechRegion+".api.cognitive.microsoft.com/speechtotext/v3.1/transcriptions");
+        request.Headers.Add("Ocp-Apim-Subscription-Key", SpeechSubscriptionKey);
+        var content = new StringContent("{\r\n\"contentUrls\": [\r\n    \"" + audio + "\"\r\n  ],\r\n  \"locale\": \"es-es\",\r\n  \"displayName\": \"My Transcription\",\r\n  \"model\": null,\r\n  \"properties\": {\r\n    \"wordLevelTimestampsEnabled\": true,\r\n    \"languageIdentification\": {\r\n      \"candidateLocales\": [\r\n        \"en-US\", \"de-DE\", \"es-ES\"\r\n      ]\r\n    }\r\n    }\r\n}", null, "application/json");
+        request.Content = content;
+        var response = await client.SendAsync(request);
+        response.EnsureSuccessStatusCode();
+        //Console.WriteLine(await response.Content.ReadAsStringAsync());
+        var responsejson = JsonConvert.DeserializeObject<dynamic>(await response.Content.ReadAsStringAsync());
+        Console.WriteLine(responsejson);
+        var output_result = responsejson.self.ToString();
+        Console.WriteLine("SELF: "+output_result);
+
+        client.Dispose();
+
+    // CALL 2: CHECK FOR FINISH
+        var client2 = new HttpClient();
+        var request2 = new HttpRequestMessage(HttpMethod.Get, output_result);
+        client2.DefaultRequestHeaders.Add("Ocp-Apim-Subscription-Key", SpeechSubscriptionKey);
+        var content2 = new StringContent(string.Empty);
+        content2.Headers.ContentType = new MediaTypeHeaderValue("application/json");
+        request2.Content = content2;
+        var response2 = await client2.SendAsync(request2);
+        response2.EnsureSuccessStatusCode();
+        //Console.WriteLine(await response2.Content.ReadAsStringAsync());
+        var responsejson2 = JsonConvert.DeserializeObject<dynamic>(await response.Content.ReadAsStringAsync());
+        Console.WriteLine(responsejson2);
+        while (responsejson2.status != "Succeeded")
+            {
+                Thread.Sleep(10000);
+                response2 = await client2.GetAsync(output_result);
+                responsejson2 = JsonConvert.DeserializeObject<dynamic>(await response2.Content.ReadAsStringAsync());
+                Console.WriteLine(responsejson2.status);
+            }
+        client2.Dispose();
+
+
+    // CALL 3: GET RESULTS URL
+
+        var client3 = new HttpClient();
+        var request3 = new HttpRequestMessage(HttpMethod.Get, output_result+"/files/");
+        request3.Headers.Add("Ocp-Apim-Subscription-Key", SpeechSubscriptionKey);
+        var content3 = new StringContent(string.Empty);
+        content3.Headers.ContentType = new MediaTypeHeaderValue("application/json");
+        request3.Content = content3;
+        var response3 = await client3.SendAsync(request3);
+        response3.EnsureSuccessStatusCode();
+        var responsejson3 = JsonConvert.DeserializeObject<dynamic>(await response3.Content.ReadAsStringAsync());
+        Console.WriteLine(responsejson3);
+        // Extract contentUrl field
+        string output_result3 = (string)responsejson3["values"][0]["links"]["contentUrl"];       
+        Console.WriteLine(output_result3);
+        client3.Dispose();
+
+    // CALL 4: GET RESULTS (TRANSCRIPTION)
+
+        var client4 = new HttpClient();
+        var request4 = new HttpRequestMessage(HttpMethod.Get, output_result3);
+        request4.Headers.Add("Ocp-Apim-Subscription-Key", SpeechSubscriptionKey);
+        var content4 = new StringContent(string.Empty);
+        content4.Headers.ContentType = new MediaTypeHeaderValue("application/json");
+        request4.Content = content4;
+        var response4 = await client4.SendAsync(request4);
+        response4.EnsureSuccessStatusCode();
+        Console.WriteLine(await response4.Content.ReadAsStringAsync());
+        var jsonObject4 = JsonConvert.DeserializeObject<JObject>(await response4.Content.ReadAsStringAsync());
+        string output_result4 = (string)jsonObject4["combinedRecognizedPhrases"][0]["lexical"];
+        Console.WriteLine(output_result4);
+        client4.Dispose();
+
+
+        //Show transcript results
+        ViewBag.Message = "TRANSCRIPTION RESULTS: \n\n"+output_result4;
+
+
+        return View("AudioTranscription", model);
+    }
+        public class SpeechToTextResponse
+    {
+        [JsonProperty("text")]
+        public string Text { get; set; }
+    }
+
+    //Upload a file to my azure storage account
+    [HttpPost]
+    public async Task<IActionResult> UploadFile(IFormFile imageFile, string prompt)
+    {
+        //Check no image
+
+        if (CheckNullValues(imageFile))
+        {
+            ViewBag.Message = "You must upload an mp3 audio file";
+            return View("AudioTranscription");
+        }
+
+        //Upload file to azure storage account
+        string url = imageFile.FileName.ToString();
+        //Console.WriteLine(url);
+        url = url.Replace(" ", "");
+        //Console.WriteLine(url);
+        BlobClient blobClient = containerClient.GetBlobClient(url);
+        var httpHeaders = new BlobHttpHeaders
+        {
+            ContentType = "audio/mpeg",
+        };
+        await blobClient.UploadAsync(imageFile.OpenReadStream(), new BlobUploadOptions { HttpHeaders = httpHeaders });
+
+        //Get the url of the file
+        Uri blobUrl = blobClient.Uri;
+
+        if (CheckImageExtension(blobUrl.ToString()))
+        {
+            ViewBag.Message = "You must upload an audio file with .mp3 extension";
+            return View("AudioTranscription", model);
+        }
+
+
+        //Call EvaluateImage with the url
+        await TranscribeAudio(blobUrl.ToString(), imageFile);
+        ViewBag.Waiting = null;
+
+        return View("AudioTranscription", model);
+    }
+
+
+
+    [ResponseCache(Duration = 0, Location = ResponseCacheLocation.None, NoStore = true)]
+    public IActionResult Error()
+    {
+        return View(new ErrorViewModel { RequestId = Activity.Current?.Id ?? HttpContext.TraceIdentifier });
+    }
+
+    private bool CheckNullValues(IFormFile imageFile)
+    {
+        if (imageFile == null)
+        {
+            return true;
+        }
+        return false;
+    }
+
+    private bool CheckImageExtension(string blobUri)
+    {
+        string uri_lower = blobUri;
+        if (uri_lower.Contains(".mp3", StringComparison.OrdinalIgnoreCase))
+        {
+            return false;
+        }
+        return true;
+    }
+}
diff --git a/src/AIHub/Models/AudioTranscriptionModel.cs b/src/AIHub/Models/AudioTranscriptionModel.cs
@@ -0,0 +1,14 @@
+namespace MVCWeb.Models;
+
+public class AudioTranscriptionModel
+{
+
+    public int? Severity { get; set; }
+    public int? Violence { get; set; }
+    public int? SelfHarm { get; set; }
+    public int? Hate { get; set; }
+    public string? Prompt { get; set; }
+    public string? Image { get; set; }
+    public string? Message { get; set; }
+
+}
diff --git a/src/AIHub/Views/AudioTranscription/AudioTranscription.cshtml b/src/AIHub/Views/AudioTranscription/AudioTranscription.cshtml
@@ -0,0 +1,72 @@
+@{
+    ViewData["Title"] = "Audio Trancription";
+}
+
+<div class="text-center">
+    <svg style="fill: var(--main-color)" xmlns="http://www.w3.org/2000/svg" height="4em"
+        viewBox="0 0 512 512"><!--! Font Awesome Free 6.4.2 by @@fontawesome - https://fontawesome.com License - https://fontawesome.com/license (Commercial License) Copyright 2023 Fonticons, Inc. -->
+        <path
+            <path d="M64 0C28.7 0 0 28.7 0 64V448c0 35.3 28.7 64 64 64H320c35.3 0 64-28.7 64-64V160H256c-17.7 0-32-14.3-32-32V0H64zM256 0V128H384L256 0zm2 226.3c37.1 22.4 62 63.1 62 109.7s-24.9 87.3-62 109.7c-7.6 4.6-17.4 2.1-22-5.4s-2.1-17.4 5.4-22C269.4 401.5 288 370.9 288 336s-18.6-65.5-46.5-82.3c-7.6-4.6-10-14.4-5.4-22s14.4-10 22-5.4zm-91.9 30.9c6 2.5 9.9 8.3 9.9 14.8V400c0 6.5-3.9 12.3-9.9 14.8s-12.9 1.1-17.4-3.5L113.4 376H80c-8.8 0-16-7.2-16-16V312c0-8.8 7.2-16 16-16h33.4l35.3-35.3c4.6-4.6 11.5-5.9 17.4-3.5zm51 34.9c6.6-5.9 16.7-5.3 22.6 1.3C249.8 304.6 256 319.6 256 336s-6.2 31.4-16.3 42.7c-5.9 6.6-16 7.1-22.6 1.3s-7.1-16-1.3-22.6c5.1-5.7 8.1-13.1 8.1-21.3s-3.1-15.7-8.1-21.3c-5.9-6.6-5.3-16.7 1.3-22.6z"/>
+    </svg>
+    <h1 class="sectionTitle">Audio Transcription</h1>
+    <p class="sectionSubTitle">Analiza tus audios usando Azure AI Speech</p>
+    <p class="sectionDetails">Sólo necesitas subir un audio (.mp3).</p>
+
+</div>
+
+@if (ViewBag.Message != null)
+{
+    <div class="row justify-content-center mt-5">
+        <div class="col-md-6">
+            <div class="alert alert-primary" role="alert">
+                @Html.Raw(ViewBag.Message.Replace("\n", "<br />"))
+            </div>
+        </div>
+    </div>
+}
+<form asp-controller="AudioTranscription" asp-action="UploadFile" method="post" enctype="multipart/form-data">
+    @* <div class="row justify-content-center mt-5">
+    <span class="form-group">
+    <label for="fname">Upload your image to analyze:</label><br>
+    <input type="text" class="form-control" id="image_url" name="image_url" value="" style="width: 70%;"/>
+    <input type="file" class="form-control-file" id="imageFile" name="imageFile" />
+    </div>
+    <button type="submit" class="btn btn-primary">Upload Image</button>
+    </div> *@
+
+    <div class="col-md-6">
+
+        <div class="form-group">
+            <label for="imageFile">Audio File:</label><br>
+            <input type="file" class="form-control-file" id="imageFile" name="imageFile" />
+            </br>
+        </div>
+        <div id="loadingPanel" style="display: none;">Transcribing...</div>
+        <button type="submit" class="btn btn-primary" onclick="submitForm()">Transcribe audio</button>
+
+    </div>
+    <script>
+        function submitForm() {
+            // Disable the button
+            var btn = document.querySelector('button[type="submit"]');
+            btn.disabled = true;
+
+            // Show the loading panel
+            var loadingPanel = document.getElementById('loadingPanel');
+            loadingPanel.style.display = 'block';
+
+            // Submit the form
+            var form = document.querySelector('form');
+            form.submit();
+        }
+        window.onload = function () {
+            // Enable the button
+            var btn = document.querySelector('button[type="submit"]');
+            btn.disabled = false;
+
+            // Hide the loading panel
+            var loadingPanel = document.getElementById('loadingPanel');
+            loadingPanel.style.display = 'none';
+        }
+    </script>
+</form>
diff --git a/src/AIHub/Views/Home/Index.cshtml b/src/AIHub/Views/Home/Index.cshtml
@@ -22,6 +22,18 @@
             </div>
         </a>
     </li>
+    <li key="audiotranscription">
+        <a asp-area="" style="text-decoration: none" asp-controller="AudioTranscription" asp-action="AudioTranscription">
+            <div class="homeBlock">
+                <svg style="fill: var(--main-color)" xmlns="http://www.w3.org/2000/svg" height="2em"
+                    viewBox="0 0 512 512"><!--! Font Awesome Free 6.4.2 by @@fontawesome - https://fontawesome.com License - https://fontawesome.com/license (Commercial License) Copyright 2023 Fonticons, Inc. -->
+                    <path d="M64 0C28.7 0 0 28.7 0 64V448c0 35.3 28.7 64 64 64H320c35.3 0 64-28.7 64-64V160H256c-17.7 0-32-14.3-32-32V0H64zM256 0V128H384L256 0zm2 226.3c37.1 22.4 62 63.1 62 109.7s-24.9 87.3-62 109.7c-7.6 4.6-17.4 2.1-22-5.4s-2.1-17.4 5.4-22C269.4 401.5 288 370.9 288 336s-18.6-65.5-46.5-82.3c-7.6-4.6-10-14.4-5.4-22s14.4-10 22-5.4zm-91.9 30.9c6 2.5 9.9 8.3 9.9 14.8V400c0 6.5-3.9 12.3-9.9 14.8s-12.9 1.1-17.4-3.5L113.4 376H80c-8.8 0-16-7.2-16-16V312c0-8.8 7.2-16 16-16h33.4l35.3-35.3c4.6-4.6 11.5-5.9 17.4-3.5zm51 34.9c6.6-5.9 16.7-5.3 22.6 1.3C249.8 304.6 256 319.6 256 336s-6.2 31.4-16.3 42.7c-5.9 6.6-16 7.1-22.6 1.3s-7.1-16-1.3-22.6c5.1-5.7 8.1-13.1 8.1-21.3s-3.1-15.7-8.1-21.3c-5.9-6.6-5.3-16.7 1.3-22.6z"/>
+                </svg>
+                <p class="homeBlockText">Audio Transcription</p>
+                <p class="homeBlockTextDetail">Transcribe audio files</p>
+            </div>
+        </a>
+    </li>
     <li key="callcenter">
         <a asp-area="" style="text-decoration: none" asp-controller="CallCenter" asp-action="CallCenter">
             <div class="homeBlock">
@@ -48,6 +60,8 @@
             </div>
         </a>
     </li>
+    </ul>
+    <ul class="homeBlockNavList">
     <li key="brandanalyzer">
         <a asp-area="" style="text-decoration: none" asp-controller="BrandAnalyzer" asp-action="BrandAnalyzer">
             <div class="homeBlock">
@@ -57,7 +71,7 @@
                         d="M36.8 192H603.2c20.3 0 36.8-16.5 36.8-36.8c0-7.3-2.2-14.4-6.2-20.4L558.2 21.4C549.3 8 534.4 0 518.3 0H121.7c-16 0-31 8-39.9 21.4L6.2 134.7c-4 6.1-6.2 13.2-6.2 20.4C0 175.5 16.5 192 36.8 192zM64 224V384v80c0 26.5 21.5 48 48 48H336c26.5 0 48-21.5 48-48V384 224H320V384H128V224H64zm448 0V480c0 17.7 14.3 32 32 32s32-14.3 32-32V224H512z" />
                 </svg>
                 <p class="homeBlockText">Brand Analyzer</p>
-                <p class="homeBlockTextDetail">Brand Analyzer: Analyze your brand's internet reputation</p>
+                <p class="homeBlockTextDetail">Analyze your brand's internet reputation</p>
             </div>
         </a>
     </li>
@@ -70,12 +84,11 @@
                         d="M320 464c8.8 0 16-7.2 16-16V160H256c-17.7 0-32-14.3-32-32V48H64c-8.8 0-16 7.2-16 16V448c0 8.8 7.2 16 16 16H320zM0 64C0 28.7 28.7 0 64 0H229.5c17 0 33.3 6.7 45.3 18.7l90.5 90.5c12 12 18.7 28.3 18.7 45.3V448c0 35.3-28.7 64-64 64H64c-35.3 0-64-28.7-64-64V64z" />
                 </svg>
                 <p class="homeBlockText">Form Analyzer</p>
-                <p class="homeBlockTextDetail">Analiza tus documentos</p>
+                <p class="homeBlockTextDetail">Summarize and ask questions to your documents</p>
             </div>
         </a>
     </li>
-</ul>
-<ul class="homeBlockNavList">
+
 
     <li key="textmodetator">
         <a asp-area="" style="text-decoration: none" asp-controller="ContentSafety" asp-action="TextModerator">