|
| 1 | +## Speech-to-Text Conversion Using Deepgram API with URLSession in Swift |
| 2 | + |
| 3 | +**Title:** Converting Pre-recorded Speech to Text Using URLSession in Swift |
| 4 | + |
| 5 | +**Code Sample:** speech-to-text/prerecorded/remote/urlsession/main.swift |
| 6 | + |
| 7 | +**Description:** This Swift code uses URLSession to send a POST request to the Deepgram API. It uploads a pre-recorded audio file (spacewalk.wav) and converts the speech in the audio file to text. The API response is then parsed and printed to the console. |
| 8 | + |
| 9 | +### speech-to-text/prerecorded/remote/urlsession/main.swift |
| 10 | + |
| 11 | +```swift |
| 12 | +import Foundation |
| 13 | + |
| 14 | +// Define the URL for the Deepgram API endpoint |
| 15 | +let url = URL(string: "https://api.deepgram.com/v1/listen")! |
| 16 | + |
| 17 | +// Define the request body |
| 18 | +let requestBody = ["url": "https://dpgr.am/spacewalk.wav"] |
| 19 | +guard let httpBody = try? JSONSerialization.data(withJSONObject: requestBody) else { |
| 20 | + print("Error: Unable to serialize request body") |
| 21 | + exit(1) |
| 22 | +} |
| 23 | + |
| 24 | +// Define the request headers |
| 25 | +var request = URLRequest(url: url) |
| 26 | +request.httpMethod = "POST" |
| 27 | +request.setValue("application/json", forHTTPHeaderField: "Content-Type") |
| 28 | +request.setValue("application/json", forHTTPHeaderField: "Accept") |
| 29 | +request.setValue("Token DEEPGRAM_API_KEY", forHTTPHeaderField: "Authorization") // Replace YOUR_DEEPGRAM_API_KEY with your actual API key |
| 30 | +request.httpBody = httpBody |
| 31 | + |
| 32 | +// Perform the HTTP request |
| 33 | +let task = URLSession.shared.dataTask(with: request) { data, response, error in |
| 34 | + guard let data = data, let httpResponse = response as? HTTPURLResponse, error == nil else { |
| 35 | + print("Error: \(error?.localizedDescription ?? "Unknown error")") |
| 36 | + return |
| 37 | + } |
| 38 | + |
| 39 | + // Check if the HTTP request was successful (status code 200) |
| 40 | + guard httpResponse.statusCode == 200 else { |
| 41 | + print("HTTP request failed with status code \(httpResponse.statusCode)") |
| 42 | + return |
| 43 | + } |
| 44 | + |
| 45 | + // Parse and print the response body |
| 46 | + if let responseBody = String(data: data, encoding: .utf8) { |
| 47 | + print("Response: \(responseBody)") |
| 48 | + } else { |
| 49 | + print("Error: Unable to parse response body") |
| 50 | + } |
| 51 | +} |
| 52 | + |
| 53 | +task.resume() |
| 54 | + |
| 55 | +// Keep the program running until the HTTP request completes |
| 56 | +RunLoop.main.run() |
| 57 | + |
| 58 | +``` |
| 59 | + |
| 60 | +## Speech-to-Text Conversion using Deepgram API with Local File |
| 61 | + |
| 62 | +**Title:** Converting Pre-recorded Audio to Text using Deepgram API |
| 63 | + |
| 64 | +**Code Sample:** speech-to-text/prerecorded/local/urlsession/main.swift |
| 65 | + |
| 66 | +**Description:** This Swift code uses the Deepgram API to convert a pre-recorded audio file into text. It reads an audio file as binary data, creates a POST request with the audio data as the body, and sends it to the Deepgram API. The API response is then printed to the console. |
| 67 | + |
| 68 | +### speech-to-text/prerecorded/local/urlsession/main.swift |
| 69 | + |
| 70 | +```swift |
| 71 | +import Foundation |
| 72 | + |
| 73 | +// Specify the URL for the Deepgram API endpoint |
| 74 | +let url = URL(string: "https://api.deepgram.com/v1/listen")! |
| 75 | + |
| 76 | +// Specify the path to the audio file |
| 77 | +let audioFilePath = "/path/to/youraudio.wav" |
| 78 | + |
| 79 | +// Read the audio file as binary data |
| 80 | +guard let audioData = FileManager.default.contents(atPath: audioFilePath) else { |
| 81 | + print("Error: Unable to read audio file") |
| 82 | + exit(1) |
| 83 | +} |
| 84 | + |
| 85 | +// Create the URLRequest object |
| 86 | +var request = URLRequest(url: url) |
| 87 | +request.httpMethod = "POST" |
| 88 | + |
| 89 | +// Set request headers |
| 90 | +request.setValue("Token DEEPGRAM_API_KEY", forHTTPHeaderField: "Authorization") // Replace YOUR_DEEPGRAM_API_KEY with your actual API key |
| 91 | +request.setValue("audio/wav", forHTTPHeaderField: "Content-Type") |
| 92 | + |
| 93 | +// Set request body with audio data |
| 94 | +request.httpBody = audioData |
| 95 | + |
| 96 | +// Create URLSession task to perform the request |
| 97 | +let task = URLSession.shared.dataTask(with: request) { data, response, error in |
| 98 | + if let error = error { |
| 99 | + print("Error: \(error)") |
| 100 | + return |
| 101 | + } |
| 102 | + |
| 103 | + guard let httpResponse = response as? HTTPURLResponse, (200...299).contains(httpResponse.statusCode) else { |
| 104 | + print("Error: Invalid response") |
| 105 | + return |
| 106 | + } |
| 107 | + |
| 108 | + if let data = data { |
| 109 | + if let responseBody = String(data: data, encoding: .utf8) { |
| 110 | + print("Response: \(responseBody)") |
| 111 | + } else { |
| 112 | + print("Error: Unable to parse response body") |
| 113 | + } |
| 114 | + } else { |
| 115 | + print("Error: No response data") |
| 116 | + } |
| 117 | +} |
| 118 | + |
| 119 | +// Start the URLSession task |
| 120 | +task.resume() |
| 121 | + |
| 122 | +// Keep the program running until the URLSession task completes |
| 123 | +RunLoop.main.run() |
| 124 | + |
| 125 | +``` |
| 126 | + |
| 127 | +## Text-to-Speech Conversion Using Deepgram API |
| 128 | + |
| 129 | +**Title:** Converting Text to Speech using URLSession in Swift |
| 130 | + |
| 131 | +**Code Sample:** text-to-speech/urlsession/main.swift |
| 132 | + |
| 133 | +**Description:** This Swift code uses the Deepgram API to convert text to speech. It sends a POST request with the text to be spoken to the Deepgram API. The response, an audio file, is then saved locally. The code uses URLSession for networking tasks. |
| 134 | + |
| 135 | +### text-to-speech/urlsession/main.swift |
| 136 | + |
| 137 | +```swift |
| 138 | +import Foundation |
| 139 | + |
| 140 | +// Specify the URL for the Deepgram API endpoint |
| 141 | +let url = URL(string: "https://api.deepgram.com/v1/speak?model=aura-asteria-en")! |
| 142 | + |
| 143 | +// Replace DEEPGRAM_API_KEY with your actual API key |
| 144 | +let apiKey = "DEEPGRAM_API_KEY" |
| 145 | + |
| 146 | +// Text to be converted to speech |
| 147 | +let textToSpeak = "Hello, how can I help you today?" |
| 148 | + |
| 149 | +// Create the URLRequest object |
| 150 | +var request = URLRequest(url: url) |
| 151 | +request.httpMethod = "POST" |
| 152 | + |
| 153 | +// Set request headers |
| 154 | +request.setValue("Token \(apiKey)", forHTTPHeaderField: "Authorization") |
| 155 | +request.setValue("application/json", forHTTPHeaderField: "Content-Type") |
| 156 | + |
| 157 | +// Create request body with text data |
| 158 | +let textData = ["text": textToSpeak] |
| 159 | +let jsonData = try! JSONSerialization.data(withJSONObject: textData) |
| 160 | +request.httpBody = jsonData |
| 161 | + |
| 162 | +// Create URLSession task to perform the request |
| 163 | +let task = URLSession.shared.dataTask(with: request) { data, response, error in |
| 164 | + if let error = error { |
| 165 | + print("Error: \(error)") |
| 166 | + return |
| 167 | + } |
| 168 | + |
| 169 | + guard let httpResponse = response as? HTTPURLResponse, (200...299).contains(httpResponse.statusCode) else { |
| 170 | + print("Error: Invalid response") |
| 171 | + return |
| 172 | + } |
| 173 | + |
| 174 | + if let audioData = data { |
| 175 | + do { |
| 176 | + // Specify the path to save the output MP3 file |
| 177 | + let outputPath = "your_output_file.mp3" |
| 178 | + try audioData.write(to: URL(fileURLWithPath: outputPath)) |
| 179 | + print("MP3 file saved at: \(outputPath)") |
| 180 | + } catch { |
| 181 | + print("Error saving MP3 file: \(error)") |
| 182 | + } |
| 183 | + } else { |
| 184 | + print("Error: No response data") |
| 185 | + } |
| 186 | +} |
| 187 | + |
| 188 | +// Start the URLSession task |
| 189 | +task.resume() |
| 190 | + |
| 191 | +// Keep the program running until the URLSession task completes |
| 192 | +RunLoop.main.run() |
| 193 | + |
| 194 | +``` |
| 195 | + |
0 commit comments