Praanscribe is an open-source Python CLI application designed to automate the segmentation of parts of speech and the annotation of sounds, as much as possible, using Praat. It uses OpenAI’s Whisper model for transcription and produces .TextGrid files compatible with Praat for linguistic analysis.
🔗 https://alicagankaya.com/praanscribe/
Automatic Speech Recognition (ASR) & Segmentation
ASR and segmentation technologies are foundational in modern phonetic and phonological research. ASR systems convert spoken input into textual or symbolic representations, making linguistic analysis more efficient (Jurafsky & Martin, 2023). Temporal segmentation—identifying boundaries at the sentence, word, or phoneme level—is key to aligning acoustic signals with linguistic units for qualitative and quantitative research.
Praat (Boersma & Weenink, 2001) is a powerful tool for acoustic analysis and manual annotation. It supports multi-tiered .TextGrid annotations but requires time-consuming manual effort, especially for large datasets.
Praanscribe addresses this challenge by integrating Whisper-based ASR with Praat, generating sentence- and word-aligned .TextGrid files automatically—streamlining research workflows and enhancing reproducibility.
- Written in Python
- Runs OpenAI’s Whisper locally
- Supports 5 model sizes:
tiny,base,small,medium, andlarge - Supports 99 languages
- Automatically generates:
- Sentence-level and word-level transcriptions
.TextGridfiles aligned with transcription timestamps- Grapheme-to-phoneme (G2P) conversion for phonological tiers
Note: As Whisper is corpus-trained, accuracy may drop with pseudowords or rare lexemes. G2P methods work well for standard forms but may not capture dialectal variation.
- Transcribe audio using Whisper (ASR)
- Extract sentence and word timestamps
- Convert data to
.TextGridformat - Align tiers in Praat based on user preferences
- Run the
praanscribescript in a Python environment. - Provide your audio input (
.wav,.flac,.mp3,.aac, etc.). - Choose the Whisper model size (
tinytolarge) based on accuracy and compute power. - Select tiers to include. Default tiers:
phoneswords-phonowordsphonoortho
- The generated
.TextGridfile will be saved in the same directory as the audio file.
These research projects have used Praanscribe and opted to be publicly listed here. I’m honored to support their work.
- Kaya, A. Ç. (2024). Ölçünlü Türkçenin ünlü formant frekansları. TÜBİTAK 2209-A.
- Uzun, İ. P. (2025). Türkçede Bileşik Sözcüklerin Sözlü Dilde Üretim Süreçlerine İlişkin Akustik Sesbilgisel Görünümler. TÜBİTAK 1002-A Hızlı Destek Modülü. (Project No: 223K318)
- Boersma, P. & Weenink, D. (2025). Praat: doing phonetics by computer [Computer program]. Version 6.4.27. http://www.praat.org/
- Goldman, J.-P. (2011). Easyalign: an automatic phonetic alignment tool under praat. Interspeech 2011, 3233–3236. https://doi.org/10.21437/Interspeech.2011-815
- Jurafsky, D., & Martin, J. H. (2023). Speech and Language Processing (3rd ed.). Draft chapters online.
⭐️ Pull requests, issues, and contributions are welcome!