jake/vOOice
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add support for scribe v2 model

Browse Source
This commit is contained in:
Beingpax 2025-11-17 10:14:32 +05:45
parent 35a08dce7b
commit 308098694d
2 changed files with 54 additions and 65 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
VoiceInk/Models/PredefinedModels.swift
View File

@ -205,14 +205,24 @@ import Foundation
), ),
CloudModel( CloudModel(
name: "scribe_v1", name: "scribe_v1",
displayName: "Scribe v1 (ElevenLabs)", displayName: "Scribe v1",
description: "ElevenLabs' Scribe model for fast and accurate transcription.", description: "ElevenLabs' Scribe model for fast & accurate transcription.",
provider: .elevenLabs, provider: .elevenLabs,
speed: 0.7, speed: 0.7,
accuracy: 0.98, accuracy: 0.98,
isMultilingual: true, isMultilingual: true,
supportedLanguages: getLanguageDictionary(isMultilingual: true, provider: .elevenLabs) supportedLanguages: getLanguageDictionary(isMultilingual: true, provider: .elevenLabs)
), ),
CloudModel(
name: "scribe_v2_realtime",
displayName: "Scribe v2",
description: "ElevenLabs' Scribe v2 model for the most accurate transcription.",
provider: .elevenLabs,
speed: 0.75,
accuracy: 0.99,
isMultilingual: true,
supportedLanguages: getLanguageDictionary(isMultilingual: true, provider: .elevenLabs)
),
CloudModel( CloudModel(
name: "nova-2", name: "nova-2",
displayName: "Nova (Deepgram)", displayName: "Nova (Deepgram)",

105
VoiceInk/Services/CloudTranscription/ElevenLabsTranscriptionService.swift
View File

@ -1,107 +1,86 @@
import Foundation import Foundation
import OSLog
class ElevenLabsTranscriptionService { class ElevenLabsTranscriptionService {
private let apiURL = URL(string: "https://api.elevenlabs.io/v1/speech-to-text")!
private let logger = Logger(subsystem: "com.prakashjoshipax.voiceink", category: "ElevenLabsTranscriptionService")
func transcribe(audioURL: URL, model: any TranscriptionModel) async throws -> String { func transcribe(audioURL: URL, model: any TranscriptionModel) async throws -> String {
let config = try getAPIConfig(for: model) guard let apiKey = UserDefaults.standard.string(forKey: "ElevenLabsAPIKey"), !apiKey.isEmpty else {
throw CloudTranscriptionError.missingAPIKey
}
let boundary = "Boundary-\(UUID().uuidString)" let boundary = "Boundary-\(UUID().uuidString)"
var request = URLRequest(url: config.url) var request = URLRequest(url: apiURL)
request.httpMethod = "POST" request.httpMethod = "POST"
request.setValue("multipart/form-data; boundary=\(boundary)", forHTTPHeaderField: "Content-Type") request.setValue("multipart/form-data; boundary=\(boundary)", forHTTPHeaderField: "Content-Type")
request.setValue("application/json", forHTTPHeaderField: "Accept") request.setValue("application/json", forHTTPHeaderField: "Accept")
request.setValue(config.apiKey, forHTTPHeaderField: "xi-api-key") request.setValue(apiKey, forHTTPHeaderField: "xi-api-key")
let body = try createElevenLabsRequestBody(audioURL: audioURL, modelName: config.modelName, boundary: boundary) let body = try createRequestBody(audioURL: audioURL, modelName: model.name, boundary: boundary)
let (data, response) = try await URLSession.shared.upload(for: request, from: body) let (data, response) = try await URLSession.shared.upload(for: request, from: body)
guard let httpResponse = response as? HTTPURLResponse else { guard let httpResponse = response as? HTTPURLResponse else {
throw CloudTranscriptionError.networkError(URLError(.badServerResponse)) throw CloudTranscriptionError.networkError(URLError(.badServerResponse))
} }
logger.notice("ElevenLabs API Response Status: \(httpResponse.statusCode)")
if let responseBody = String(data: data, encoding: .utf8) {
logger.notice("ElevenLabs API Response Body: \(responseBody)")
}
if !(200...299).contains(httpResponse.statusCode) { if !(200...299).contains(httpResponse.statusCode) {
let errorMessage = String(data: data, encoding: .utf8) ?? "No error message" let errorMessage = String(data: data, encoding: .utf8) ?? "No error message"
throw CloudTranscriptionError.apiRequestFailed(statusCode: httpResponse.statusCode, message: errorMessage) throw CloudTranscriptionError.apiRequestFailed(statusCode: httpResponse.statusCode, message: errorMessage)
} }
do { do {
let transcriptionResponse = try JSONDecoder().decode(TranscriptionResponse.self, from: data) let transcriptionResponse = try JSONDecoder().decode(ElevenLabsTranscriptionResponse.self, from: data)
return transcriptionResponse.text return transcriptionResponse.text
} catch { } catch {
throw CloudTranscriptionError.noTranscriptionReturned throw CloudTranscriptionError.noTranscriptionReturned
} }
} }
private func getAPIConfig(for model: any TranscriptionModel) throws -> APIConfig { private func createRequestBody(audioURL: URL, modelName: String, boundary: String) throws -> Data {
guard let apiKey = UserDefaults.standard.string(forKey: "ElevenLabsAPIKey"), !apiKey.isEmpty else {
throw CloudTranscriptionError.missingAPIKey
}
guard let apiURL = URL(string: "https://api.elevenlabs.io/v1/speech-to-text") else {
throw NSError(domain: "ElevenLabsTranscriptionService", code: -1, userInfo: [NSLocalizedDescriptionKey: "Invalid API URL"])
}
return APIConfig(url: apiURL, apiKey: apiKey, modelName: model.name)
}
private func createElevenLabsRequestBody(audioURL: URL, modelName: String, boundary: String) throws -> Data {
var body = Data() var body = Data()
let crlf = "\r\n"
guard let audioData = try? Data(contentsOf: audioURL) else { body.append(formField: "file", fileName: audioURL.lastPathComponent, fileData: try Data(contentsOf: audioURL), mimeType: "audio/wav", boundary: boundary)
throw CloudTranscriptionError.audioFileNotFound body.append(formField: "model_id", value: modelName, boundary: boundary)
} body.append(formField: "temperature", value: "0.0", boundary: boundary)
body.append(formField: "tag_audio_events", value: "false", boundary: boundary)
// File
body.append("--\(boundary)\(crlf)".data(using: .utf8)!)
body.append("Content-Disposition: form-data; name=\"file\"; filename=\"\(audioURL.lastPathComponent)\"\(crlf)".data(using: .utf8)!)
body.append("Content-Type: audio/wav\(crlf)\(crlf)".data(using: .utf8)!)
body.append(audioData)
body.append(crlf.data(using: .utf8)!)
// Model ID
body.append("--\(boundary)\(crlf)".data(using: .utf8)!)
body.append("Content-Disposition: form-data; name=\"model_id\"\(crlf)\(crlf)".data(using: .utf8)!)
body.append(modelName.data(using: .utf8)!)
body.append(crlf.data(using: .utf8)!)
// Disable audio event tagging
body.append("--\(boundary)\(crlf)".data(using: .utf8)!)
body.append("Content-Disposition: form-data; name=\"tag_audio_events\"\(crlf)\(crlf)".data(using: .utf8)!)
body.append("false".data(using: .utf8)!)
body.append(crlf.data(using: .utf8)!)
body.append("--\(boundary)\(crlf)".data(using: .utf8)!)
body.append("Content-Disposition: form-data; name=\"temperature\"\(crlf)\(crlf)".data(using: .utf8)!)
body.append("0".data(using: .utf8)!)
body.append(crlf.data(using: .utf8)!)
let selectedLanguage = UserDefaults.standard.string(forKey: "SelectedLanguage") ?? "auto" let selectedLanguage = UserDefaults.standard.string(forKey: "SelectedLanguage") ?? "auto"
if selectedLanguage != "auto", !selectedLanguage.isEmpty { if selectedLanguage != "auto", !selectedLanguage.isEmpty {
body.append("--\(boundary)\(crlf)".data(using: .utf8)!) body.append(formField: "language_code", value: selectedLanguage, boundary: boundary)
body.append("Content-Disposition: form-data; name=\"language_code\"\(crlf)\(crlf)".data(using: .utf8)!)
body.append(selectedLanguage.data(using: .utf8)!)
body.append(crlf.data(using: .utf8)!)
} }
body.append("--\(boundary)--\(crlf)".data(using: .utf8)!) body.append("--\(boundary)--\r\n".data(using: .utf8)!)
return body return body
} }
private struct APIConfig { private struct ElevenLabsTranscriptionResponse: Decodable {
let url: URL let text: String
let apiKey: String }
let modelName: String }
private extension Data {
mutating func append(formField: String, value: String, boundary: String) {
let crlf = "\r\n"
append("--\(boundary)\(crlf)".data(using: .utf8)!)
append("Content-Disposition: form-data; name=\"\(formField)\"\(crlf)\(crlf)".data(using: .utf8)!)
append(value.data(using: .utf8)!)
append(crlf.data(using: .utf8)!)
} }
private struct TranscriptionResponse: Decodable { mutating func append(formField: String, fileName: String, fileData: Data, mimeType: String, boundary: String) {
let text: String let crlf = "\r\n"
let language: String? append("--\(boundary)\(crlf)".data(using: .utf8)!)
let duration: Double? append("Content-Disposition: form-data; name=\"\(formField)\"; filename=\"\(fileName)\"\(crlf)".data(using: .utf8)!)
let x_groq: GroqMetadata? append("Content-Type: \(mimeType)\(crlf)\(crlf)".data(using: .utf8)!)
append(fileData)
struct GroqMetadata: Decodable { append(crlf.data(using: .utf8)!)
let id: String?
}
} }
} }