Ensuring the best quality transcription with your upload
While Konch provides a top rated, cost effective, high quality A.I. transcriptions, the world of artificial intelligence still has a long way to go to interpret audio as well as a human. This is the case whether it be from multi-billion dollar companies like Google, Amazon, or Microsoft providing automated speech recognition (ASR) services, or academic research on the technology.
In order to give our ASR models the best chance of getting accurate results, here are some tips on how to record and what audio you upload transcribes optimally.
- Though it can often transcribe well, avoid background music when possible to avoid issues
- Use a higher quality microphone or keep the microphone close to the speaker.
- Coach those in your recording to avoid speaking over one another. Even with human transcriptions, two or more people talking at the same time can be hard to understand.
- Background hiss or static can create confusion with A.I. models. Avoid fans running in the background, or any other background noise when possible.
If you can not control the above variables, and our ASR model is unable to transcribe words in your audio properly, Konch provides our "Precision" product that pipes the A.I. results to a crown of certified "Correctors" that run through the results and ensures speaker segmentation and transcription is human-grade quality at 99% accuracy at a far less expensive rate than human transcription services can offer alone. Additionally, all our "Draft" (pure A.I. based transcriptions) and our "Precision" (Human augmented review) result in time-coded words and sections that can be converted to subtitles, annotation logs, word docs, and more.