Bitmanager strengthened its speech AI research capabilities in the United Arab Emirates. One of the studies presented at Interspeech 2026 examined AI-generated speech detection across 66 languages.
Newszy: Dubai, United Arab Emirates – October 2026. Bitmanager researchers Kirill Borodin and Vasily Kudryavtsev presented scientific papers at Interspeech 2026 in Sydney. Interspeech was the flagship conference of the International Speech Communication Association (ISCA) and the world’s largest scientific conference dedicated to speech processing technologies.
Acceptance into the main programme marked an important step in the development of Bitmanager’s research capabilities. Multilingual speech technologies were among the company’s strategic priorities.
One of the presented studies examined the detection of AI-generated speech and voice deepfakes across languages. The authors developed LRLspoof, a multilingual research dataset containing more than 2,700 hours of synthetic speech in 66 languages, generated using 24 speech synthesis systems.
The findings showed that synthetic speech detection performance could vary substantially from one language to another. Strong results in one language did not necessarily translate into equally reliable performance in another.
Developing high-quality multilingual speech AI required local data, research, model training and adaptation, and engineering expertise. Bitmanager outlined plans to build these capabilities in the United Arab Emirates, working on speech recognition and synthesis, voice AI agents, speech analytics, and infrastructure for running voice models in real time.
Other presented papers involving Bitmanager researchers addressed speech data preparation and the behaviour of automatic speech recognition systems. These included errors in Whisper models that caused transcriptions to contain text that was not spoken in the original recording.
“Our goal is to ensure that businesses operating across languages are fully supported by modern AI technologies. Achieving this requires our own data, models, research and local expertise. Acceptance of our papers at Interspeech demonstrates the scientific calibre of the team. We welcome collaboration with universities and companies on research and development, speech data preparation and the testing of speech AI models,” said Vladimir Fiseyskii, founder and CEO of Bitmanager.
About Bitmanager
Bitmanager operates in the United Arab Emirates. The company develops voice AI technologies for businesses. Its work spans voice AI agents, speech analytics, speech recognition and synthesis, audio processing, and multilingual speech technologies.