French AI startup Gladia aims to revolutionize the way companies interact with audio data with its advanced audio transcription application programming interface (API). The company’s API is designed to integrate seamlessly with other products and offers superior performance compared to existing solutions. By leveraging Gladia’s technology foundation, new possibilities and use cases around audio can be unlocked.
Current audio transcription APIs offered by major cloud providers like Google, Amazon, and Microsoft have limitations that Gladia aims to address. Firstly, existing APIs are expensive, with the cost of transcribing an hour of audio typically ranging from $1.50 to $2. Secondly, the output quality can be unreliable, especially for languages with limited support. Additionally, transcription APIs are often slow, taking over 15 minutes to transcribe an hour of audio, which may not be suitable for industries requiring real-time transcriptions.
Gladia is built on Whisper, OpenAI’s open-source transcription model. While Whisper serves as a solid foundation, Gladia has optimized and enhanced it to ensure fast and accurate transcriptions. Whisper, which incorporates elements of GPT-2, tends to produce hallucinations and overly frequent common phrases. Gladia has invested significant effort in mitigating these issues and improving the model’s performance. The company has also implemented pre-processing and post-processing algorithms to further enhance the transcription results.
Promising exceptional performance, Gladia offers audio transcriptions at a significantly reduced cost of $0.61 per hour, with a transcription process that takes approximately 60 seconds. Its API features include speaker detection, timestamps, language detection, automatic punctuation, and casing. The API supports various formats, including JSON, SRT, and VTT, catering to companies seeking subtitle generation capabilities.
Initial feedback on Gladia’s API has been highly positive, with users praising its accuracy and efficiency compared to competitors like Google and Azure. Gladia has already gained traction with call center companies, virtual meeting services, and video publishers, partnering with Claap, Livestorm, and Selectra.
Gladia recently secured $4 million in seed funding led by New Wave, with participation from investors such as Sequoia, Cocoa, and prominent business angels. The company’s vision extends beyond audio transcription, aiming to develop additional features on top of its robust technical foundation. These features may include translation, content summarization, topic categorization, automated chapter creation, and sentiment analysis, enabling companies to extract valuable insights from audio content.
As Gladia continues to evolve, its long-term goal is to enhance audio data from a two-dimensional perspective to a more comprehensive three-dimensional intelligence, augmenting the capabilities and possibilities of transcription services. With its commitment to innovation, Gladia envisions transcription becoming a commodity while the added options and intelligence it provides become increasingly vital to customers.