Case Study: OpenAI accelerates speech-to-text pre-annotation with SuperAnnotate

OpenAI powers Whisper with SuperAnnotate to transcribe 100 languages and 680,000 hours of audio

OpenAI developed Whisper, a powerful automatic speech recognition (ASR) system capable of transcribing and translating audio in many languages. The challenge was that Whisper, while technologically advanced, was only released as code and pre-trained models, making it inaccessible to users without programming expertise who needed an easy way to utilize the technology.

To address this, SuperAnnotate created a user-friendly Jupyter notebook that can be run on Google Colab, eliminating the need for coding skills. This notebook allows users to easily apply the Whisper model to their own audio files. Furthermore, SuperAnnotate integrated a specific workflow where Whisper's transcriptions are uploaded to its platform as pre-annotations, significantly speeding up the process of creating new, accurately labeled ASR datasets by allowing annotators to simply correct the AI's output rather than start from scratch.

OpenAI