AI Language Accuracy at Happy Scribe

AI Quality: AI-Powered, Human perfected

At Happy Scribe we believe you deserve to know the quality of the AI that you are buying.

We check our model against several competitors and are fully transparent with the results. The data comes from the corrections that our professional linguists make on top of each automatic output. Above, you can see the real-time results of our test: a comparison of the word-error-rate of our model, and three industry-standard competitor models.

On the off-chance that our model isn't the best, we’ll use a competitor’s model. No ego. No fuss. Just the best quality output for you.

“The accuracy of Happy Scribe’s automatic transcription is why we opted to use them“ - Georgie Nicholls, Production Manager, TwoFour

How Our AI Works

AI transcription is the process of using artificial intelligence to convert audio or video recordings into text. At Happy Scribe our models are trained on over 200 years worth of recordings, which means they are highly accurate across a wide range of languages, voices and accents.

AI Transcription Accuracy: Word-Error-Rate

AI transcription accuracy is measured via the metric: Word-Error-Rate. Put simply, WER is the ratio of errors in a transcript to the total words spoken. A lower WER in speech-to-text means better accuracy in recognizing speech. For example, a 20% WER means the transcript is 80% accurate. Out in the real world, both the WER formula and its applicability are a little more nuanced.

For example, if I have an AI transcription with 100 words, that is reviewed by a linguist who finds that 5 words were not captured correctly by the AI. This transcription would have a word-accuracy rate of 95/100 or 95%.

Why Is AI Accuracy Important?

As the AI accuracy improves, you save time editing and perfecting your work. Higher quality output from the AI means you can spend more of your time creating the content you want to create, rather than fixing mistakes. Whether you use an internal team or external resources your team can do more with less as AI language accuracy improves.