Comparative Analysis of Energy Consumption and Carbon Footprint in Automatic Speech Recognition Systems

Strategic Insights -

Researchers from Abdelmalek Essaadi University in Morocco have conducted a study to investigate the energy consumption and carbon footprint of two prominent automatic speech recognition (ASR) systems: OpenAI's Whisper and Google's Speech-to-Text API. The study utilized a public Kaggle dataset of 20,000 short audio clips in Urdu and evaluated both local and cloud-based speech recognition approaches using CodeCarbon, PyJoule, and PowerAPI for comprehensive energy profiling.

Key Takeaways:

  • The study found that the cloud-based solution (Google's Speech-to-Text API) showed substantially lower environmental impact compared to the local solution (OpenAI's Whisper) despite comparable accuracy.
  • The researchers evaluated both systems using a public Kaggle dataset of 20,000 short audio clips in Urdu, demonstrating the feasibility of using public datasets for energy profiling.
  • The study highlighted the importance of considering the environmental impact of AI deployment, particularly in areas where energy consumption is high.
  • The researchers discussed the implications of their findings for sustainable AI deployment and minimizing the ecological footprint of speech recognition technologies.
  • The study used CodeCarbon, PyJoule, and PowerAPI for comprehensive energy profiling, demonstrating the effectiveness of energy profiling tools in assessing the environmental impact of ASR systems.
  • The authors, Jalal El Bahri, Mohamed Kouissi, and Mohammed Achkari Begdouri, emphasized the need for more research on the energy consumption and carbon footprint of ASR systems.
  • The study's findings have significant implications for the development of sustainable AI technologies.

Statistics:

  • The study used a public Kaggle dataset of 20,000 short audio clips in Urdu.
  • The researchers evaluated both local and cloud-based speech recognition approaches using CodeCarbon, PyJoule, and PowerAPI.
  • The study found that the cloud-based solution showed 30.6% lower energy consumption compared to the local solution.
  • The research concluded that the cloud-based solution had a 25.1% lower carbon footprint compared to the local solution.

Sources:

  • Research: Comparative Analysis of Energy Consumption and Carbon Footprint in Automatic Speech Recognition Systems: A Case Study Comparing Whisper and Google Speech-to-Text. Computer Sciences & Mathematics Forum, 2025, 10(1):6-0.
  • Publisher: MDPI AG.
  • Journal article: https://doi-org.sdpl.idm.oclc.org/10.3390/cmsf2025010006.
  • Contact information: Jalal El Bahri, SIGL Research Laboratory, ENSATE of Tetouan, Abdelmalek Essaadi University, Tetouan 93040, Morocco.
  • Additional authors: Mohamed Kouissi, Mohammed Achkari Begdouri.
  • Keywords: Abdelmalek Essaadi University, Tetouan, Morocco, Africa, Cloud Computing, Information Technology.