The Finalist | Transcriber
Even though life seems to be slowly returning to normality for most of us, as the COVID pandemic is passing its critical stage. However, a grand majority of people gatherings such as public forums or even cultural events like theatre face the challenge of going remote. For most of us, joining a live theatre session or debate is as simple as joining a video call. If you analyse the crowd with auditory impairment, this trivial task is no longer as trivial.
The platform offers webinar-style meetings, where one main host broadcasts audio and video to the participants whilst they interact via a real-time chat private to that meeting, for example when asking questions. The platform provides a live transcript generation service by using the presenter’s audio stream to make inferences on a deep learning engine powered by the Mozilla DeepSpeech library. The model supports only English for now but it can be trained for other languages, such as Polish.
We value your privacy – unlike some cloud conferencing providers – all meetings are one-time occurrences. When the presenter ends the conversation, no audio transcripts are stored on the server-side, everything is up to the participants. The audio from the presenter is streamed directly in the backend and chunked into a configurable amount of seconds. The raw audio bytes are transferred by a secure channel to the transcription service. When the transcript becomes available (via async communication and message queues), the UIs of the participants are updated with the live transcript of the person speaking, allowing interaction with people that might have hearing disabilities or just with your regular forum, as the transcript functionality could
As a business model, we provide free services to educational or non-profit organisations, leveraging the disability-easing side of things but we charge enterprises that would want to use our service to generate automatic meeting notes, for example. We envision a world of the future, where people can interact digitally with the same level of autonomy as in the real world.
Team name: Transcribers
Project name: Transcriber
Team members: Cristian Schuszter, Kamil Faber, Karolina Cynk