Tinkoff's voice assistant Oleg now available in Clubhouse

Tinkoff has integrated its assistant Oleg into Clubhouse, making it the first voice assistant, speech recognition and synthesis solution available in this audio-chat social network.

Oleg will be a full-fledged user helping room creators to communicate and moderate discussions in Clubhouse utilising its text-to-speech and speech-to-text capabilities (Tinkoff VoiceKit) in real time.

Tinkoff’s voice assistant will be able to enter rooms, transcribe speech in real time, and stream the text in his Oleg in the Clubhouse Telegram channel. He can also moderate Clubhouse rooms, voice questions to speakers, remind users about time limits, regulations, etc.

Oleg made his debut appearance in Clubhouse on 11 March converting speech to text and streaming the results from the Tinkoff Investments room where Oliver Hughes and other of the Group’s senior executives held a conference call for investors and journalists. The room was created to discuss Tinkoff Group’s financial performance and record net profit in 2020.

Pavel Kalaidin, Director of Artificial Intelligence at Tinkoff

Our voice assistant team is currently experimenting with various user scenarios in Clubhouse to determine how room creators or listeners can benefit from our technologies.

We have already successfully tested Oleg’s ability to transcribe audio calls in real time streaming them in his Telegram channel. The feature was piloted in the Clubhouse room created to discuss Tinkoff’s 2020 financial results.

Oleg can also come in handy when listeners are unable to voice a question to speakers, for example when it is too noisy or they do not want to interrupt them. For such cases, we are designing an interface through which users can forward their questions to Oleg’s Telegram chat. Oleg will then voice the question with perfect pronunciation, keeping the user anonymous, if necessary.

One of the challenges in group speech recognition is the summarisation of information. Interjections, fillers and incoherent speech make reading even a good transcript difficult. For that reason, we are looking into ways of processing the text and capturing the gist of what is said to create a shorter and more readable transcript.

We are open to working with Clubhouse communities to make our voice assistant a useful tool for content makers and listeners.

Voice assistant Oleg relies on Tinkoff VoiceKit, a set of proprietary speech-to-text and text-to-speech technologies.

Tinkoff VoiceKit features deep neural network models for speech recognition and synthesis developed by Tinkoff over the recent years as part of its AI First strategy and used to create Oleg, the world’s first proprietary financial voice assistant.

It can be used to:

  • create voice assistants;
  • create robots to automate call centres;
  • accelerate production of audiobooks and voice-overs and speed up video editing;
  • build a speech analytics system based on transcripts (e.g. to supervise operators at a call centre);
  • create applications for people with disabilities;
  • transcribe any public speeches recorded on audio;
  • facilitate SEO activities and full-text search of audio and video recordings.

Latest news