Convert MP3 to XLS

A legacy office system that accepts only Excel 97–2003 files cannot use an MP3 recording as spreadsheet data. Transcribe the recording first, then save the timestamped transcript as .xls for filtering, editing, review, or archival.

What MP3 is

MP3 is a lossy compressed audio format used for interviews, meetings, lectures, podcasts, music, telephone recordings, and voice notes. Smartphones, digital recorders, media players, audio editors such as Audacity, and transcription applications can create or save MP3 files.

An MP3 contains encoded sound, not selectable words or worksheet cells. A conventional file converter can change a container or encoding, but it cannot produce a meaningful spreadsheet from speech without automatic speech recognition or manual transcription.

What XLS is

XLS is the binary spreadsheet format used by Microsoft Excel 97–2003. It stores worksheets containing rows, columns, text, formulas, formatting, and other spreadsheet data. A transcript can use columns such as Start, End, Speaker, and Transcript, allowing sorting, filtering, correction, and import into legacy office systems.

The format’s main advantage is compatibility with software that rejects XLSX. XLS has lower limits than XLSX, including a maximum of 65,536 rows and 256 columns per worksheet, and it does not preserve all newer Excel features.

How to perform the conversion

Desktop method with Whisper and LibreOffice

1. Install FFmpeg, Python, OpenAI Whisper, and LibreOffice. FFmpeg supplies audio decoding; Whisper transcribes speech; LibreOffice Calc converts the resulting table to legacy XLS.
2. Install the Python packages:

pip install -U openai-whisper srt

3. Run Whisper against the recording. Replace en with the recording’s language code:

whisper recording.mp3 --model medium --language en --task transcribe --output_format srt --output_dir transcript

Whisper normally writes transcript/recording.srt. Larger models can improve recognition but require more memory and processing time. Supplying the correct language avoids unreliable automatic language detection.

4. Save this script as srt_to_csv.py:

import csv
import srt

with open('transcript/recording.srt', encoding='utf-8') as source:
    subtitles = list(srt.parse(source.read()))

with open('transcript.csv', 'w', newline='', encoding='utf-8-sig') as target:
    writer = csv.writer(target)
    writer.writerow(['Start', 'End', 'Speaker', 'Transcript'])
    for subtitle in subtitles:
        writer.writerow([
            str(subtitle.start),
            str(subtitle.end),
            '',
            subtitle.content.replace('\n', ' ')
        ])

5. Run the script from the folder containing it:

python srt_to_csv.py

6. Open transcript.csv in LibreOffice Calc. In the import dialog, select UTF-8 and comma as the separator.
7. Choose File → Save As, select Excel 97–2003 (.xls), and accept the format warning. The CSV is an intermediate table; the saved XLS is the requested deliverable.

Online method

Upload the MP3 to a transcription service such as Happy Scribe, Sonix, or Trint, select the language, review the generated transcript, and export CSV or XLSX if available. Open the export in Excel or LibreOffice and use File → Save As to create .xls when the service does not provide legacy XLS directly. Export options change, so verify that the chosen service currently supports spreadsheet output.

CloudConvert and similar file-conversion services can convert CSV or XLSX to XLS, but they do not transcribe MP3 audio. Do not upload confidential, regulated, or sensitive recordings unless the service’s storage, processing, and deletion terms meet your requirements.

Quality and compatibility limits

• Recognition accuracy depends on recording quality, background noise, accents, microphone placement, overlapping speech, language, and specialist vocabulary. Check names, numbers, dates, punctuation, and technical terms against the audio.
• Standard Whisper output includes timestamps but does not reliably identify speakers. The blank Speaker column must be completed manually, or a diarization workflow such as WhisperX can generate speaker labels for review.
• Each subtitle segment becomes one spreadsheet row in this workflow. Long recordings can exceed the XLS row limit; split the transcript across worksheets or files when necessary.
• Use XLSX as the working format when possible and create XLS only for the legacy application that requires it. Saving to XLS can remove newer formulas, formatting, and other features.
• Embedding an MP3 object in an XLS workbook only attaches audio for playback; it does not convert speech into searchable spreadsheet text.