Elamur Research⌕
☰
home / papers / 2026 / Software for Automatic Speech Recognition via Whisper models applied to Oral History interviews in the Portuguese language

Software for Automatic Speech Recognition via Whisper models applied to Oral History interviews in the Portuguese language

Paper · 2026 · Edgleide de Oliveira Clemente da Silva, Fernando Rezende Zagatti, Filipe Loyola Lopes et al.

Summary

Oral History projects produce long interview recordings, and transcribing them by hand is slow and tiring. Commercial transcription services usually process audio in the cloud, which conflicts with the confidentiality these interviews often require. The authors built Ethos AT, a Python desktop application that runs OpenAI Whisper models entirely on the user's own computer. It targets researchers without programming skills. The tool accepts common audio and video formats, converts them to MP3 with FFmpeg, and can split long files into user-defined chunks. It then exports plain-text and timestamped transcripts. The authors also compared six Whisper variants on Portuguese Oral History interviews, looking at accuracy (WER, CER), error types and processing time. Larger models were the most lexically faithful but slowest. Smaller ones ran faster but struggled with proper names, acronyms and technical terms. The turbo model came close to the large models in accuracy while being much faster. The work matters because it gives humanities researchers a private, easy-to-use transcription tool, though human review of the output is still needed.

Key points

  • Ethos AT (ethos-at.github.io) has a modular pipeline with four stages: media import and conversion, segmentation by chunk size, Whisper inference, and export.
    • It produces two output files: transcription_full.txt and transcription_full_timestamp.txt.
  • Supported inputs are .mp4, .mkv, .avi, .mov, .mp3 and .wav. The interface offers Whisper variants from tiny to turbo.
  • Evaluated models were small, medium, large, large-v2, large-v3 and turbo.
    • Turbo has 809M parameters, needs about 6 GB of VRAM and runs about 8x faster than the large models (1550M parameters, about 10 GB).
  • Recommendations: use turbo or large-v3 when accuracy matters most, and small or medium when speed matters more than errors.
  • Known issues and future plans:
    • Testing revealed an occasional "[WinError 5] Access denied" error when writing temporary chunks into folders with restricted permissions or OneDrive sync.
    • Planned additions are speaker diarization and institutional glossaries for acronyms and technical terms.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 works

🧭 Topics 1Research subjects and languages the work is about.Speech & spoken language · 480 works

Cite

Cite as Edgleide de Oliveira Clemente da Silva, Fernando Rezende Zagatti, Filipe Loyola Lopes, Anderson Dias Duarte, Rodrigo Bonacin and Angela Maria Alves. 2026. Software for Automatic Speech Recognition via Whisper models applied to Oral History interviews in the Portuguese language. Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 1.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{silva2026software,
  title = {{Software for Automatic Speech Recognition via Whisper models applied to Oral History interviews in the Portuguese language}},
  author = {Edgleide {de Oliveira Clemente da Silva}, Fernando Rezende Zagatti, Filipe Loyola Lopes, Anderson Dias Duarte, Rodrigo Bonacin and Angela Maria Alves},
  booktitle = {Proceedings of the 17th International Conference on Computational Processing of Portuguese (PROPOR 2026) - Vol. 1},
  year = {2026},
  url = {https://aclanthology.org/2026.propor-1.30/}
}

References

20 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Fabio Battaglia. 2024. From “listen and repeat” to “listen and revise”: how to transcribe interviews offline quickly and for free using voice recognition software. International Journal of Qualitative Methods, 23:16094069241247473. find
  2. Christian Bokhove and Christopher Downey. 2018. Automated generation of ‘good enough’transcripts as a first step to transcription of audio-recorded data. Methodological innovations, 11(2):2059799118790743. find
  3. Arnaldo Candido Junior, Edresson Casanova, Anderson Soares, Frederico Santos de Oliveira, Lucas Oliveira, Ricardo Corso Fernandes Junior, Daniel Peixoto Pinto da Silva, Fernando Gorgulho Fayet, Bruno Baldissera Carlotto, Lucas Rafael Stefanel Gris, and 1 others. 2023. Coraa asr: a large corpus of sp… find
  4. Mohsen Fatehifar, Josef Schlittenlacher, Ibrahim Almu-farrij, David Wong, Tim Cootes, and Kevin J Munro. 2025. Applications of automatic speech recognition and text-to-speech technologies for hearing assessment: a scoping review. International Journal of Audiology, 64(6):537–548. find
  5. Sanchit Gandhi, Patrick Von Platen, and Alexander M Rush. 2023. Distil-whisper: Robust knowledge distillation via large-scale pseudo labelling. arXiv preprint arXiv:2311.00430. — [arXiv:2311.00430](https://arxiv.org/abs/2311.00430) find
  6. Howard B. Goodman and Pam Rowland. 2021. Deficiencies of compliancy for data and storage. In National Cyber Summit (NCS) Research Track 2020, pages 170–192, Cham. Springer International Publishing. find
  7. Michael Gref. 2022. Robust Speech Recognition via Adaptation for German Oral History Interviews. Ph.D. thesis, Universitäts-und Landesbibliothek Bonn. find
  8. Lucas Rafael Stefanel Gris, Arnaldo Candido Junior, Vinícius G dos Santos, Bruno A Papa Dias, Marli Quadros Leite, Flaviane Romani Fernandes Svartman, and Sandra Aluísio. 2022. Bringing nurc/sp to digital life: the role of open-source automatic speech recognition models. arXiv preprint arXiv:2210.07… — [arXiv:2210.07852](https://arxiv.org/abs/2210.07852) find
  9. Lucas Rafael Stefanel Gris, Ricardo Marcacini, Arnaldo Candido Junior, Edresson Casanova, Anderson Soares, and Sandra Maria Aluísio. 2023. Evaluating openai’s whisper asr for punctuation prediction and topic modeling of life histories of the museum of the person. arXiv preprint arXiv:2305.14580. — [arXiv:2305.14580](https://arxiv.org/abs/2305.14580) find
  10. Naoyuki Kanda, Xiong Xiao, Yashesh Gaur, Xiaofei Wang, Zhong Meng, Zhuo Chen, and Takuya Yoshioka. 2022. Transcribe-to-diarize: Neural speaker diarization for unlimited number of speakers using end-to-end speaker-attributed asr. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech… find
  11. Eduardo Medeiros, Leonel Corado, Luís Rato, Paulo Quaresma, and Pedro Salgueiro. 2023. Domain adaptation speech-to-text for low-resource european portuguese using deep learning. Future Internet, 15(5):159. find
  12. Pedro Nascimento, João C Ferreira, and Fernando Batista. 2024. Automatic transcription system for parliamentary debates in the context of assembly of the republic of portugal. International Journal of Speech Technology, 27(3):613–635. find
  13. Leontien Pragt, Peter van Hengel, Dagmar Grob, and Jan-Willem A Wasmann. 2022. Preliminary evaluation of automated speech recognition apps for the hearing impaired and deaf. Frontiers in Digital Health, 4:806076. find
  14. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust speech recognition via large-scale weak supervision. In International conference on machine learning, pages 28492–28518. PMLR.
  15. Antonio Joaquim Severino. 2016. História oral como arte da escuta, de alessandro portelli. EccoS–Revista Científica, (41):238–243. find
  16. Xian Shi, Yanni Chen, Shiliang Zhang, and Zhijie Yan. 2022. Achieving timestamp prediction while recognizing with non-autoregressive end-to-end asr model. In National Conference on Man-Machine Speech Communication, pages 89–100. Springer. find
  17. Linda Shopes. 2002. Making sense of oral history. History matters: The US survey course on the web, 6. find
  18. Radina Stoykova, Kyle Porter, and Thomas Beka. 2024. The ai act in a law enforcement context: The case of automatic speech recognition for transcribing investigative interviews. Forensic Science International: Synergy, 9:100563. find
  19. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems, 30.
  20. Armin Haberl, Jürgen Fleiß, Dominik Kowald, and Stefan Thalmann. 2024. Take the atrain. introducing an interface for the accessible transcription of interviews. Journal of Behavioral and Experimental Finance, 41:100891. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/silva-2026-software-automatic-speech-recognition-whisper/