How to remove text formatting from the segments?

[zaidbren]

Hello everyone, I am trying to use the package for transcripting audio transcription based editing, and I need the segments shown with the timestamps, however, I am trying to access the result.segment text, and its giving me result with formatted like <starttime/></0.33/> etc.

private func transcribeAudio(
        at audioURL: URL,
        whisperKit: WhisperKit,
        label: String,
        fileIndex: Int,
        totalFiles: Int
    ) async throws {
        
        let options = DecodingOptions(
            task: .transcribe,
            language: selectedLanguage.code,
            wordTimestamps: true,
            supressTokens: nil
        )
        
        let transcriptionResults: [TranscriptionResult] = try await whisperKit.transcribe(
            audioPath: audioURL.path,
            decodeOptions: options,
            callback: progressCallback
        )
        
        // Process and structure the results
        for result in transcriptionResults {
            for segment in result.segments {
                let words = (segment.words ?? []).map { wordTiming in
                    TranscriptionWord(
                        word: wordTiming.word,
                        start: Double(wordTiming.start),
                        end: Double(wordTiming.end)
                    )
                }
                
                let transcriptionSegment = TranscriptionSegment(
                    id: segment.id,
                    start: Double(segment.start),
                    end: Double(segment.end),
                    text: segment.text,
                    words: words
                )
                
                transcriptionSegments.append(transcriptionSegment)
            }
        }
        
        print("Processed \(transcriptionSegments.count) segments from \(label)")
    }

I only want to show pure text without any or any formattings

[ZachNagengast]

Hi @zaidbren, best way to do this is with the skipSpecialTokens decoding option, like this:

        let options = DecodingOptions(
            task: .transcribe,
            language: selectedLanguage.code,
            skipSpecialTokens: true,
            wordTimestamps: true,
            supressTokens: nil
        )

By default it is false

WhisperKit/Sources/WhisperKit/Core/Configurations.swift

Line 195 in 2870b46

skipSpecialTokens: Bool = false,