Podcast transcription
How does it work ?
Our platform uses the OpenAI "Whisper" model to automatically create transcripts from podcast audio.
Machine learning technology applies neural network models to audio content to detect speech and convert it into text.
Models are trained to understand a specific language and dialect. Some language have much better training sets available, and so produces much better accuracy. Accuracy is typically measured in "word error rate" (WER), where a WER of 5-10% is considered to be good quality, 20% is acceptable and 30% or more deliver relatively poort quality.
The top-performing languages for Whisper transcription accuracy are English , Italian , German , and Spanish . Mid-performing languages include French, Portuguese, and Japanese, while the worst-performing languages are Arabic and Hindi.
See this article for more details.
Why use it ?
Attach a transcript to make subtitles available to your audience, a great accessibility feature.
We also include