Do you also remember how inaccurately Siri understood our voice commands at the beginning? There’s been a lot going on in the field of automatic voice recognition in recent years. This is also shown by the current study on the accuracy of Google Assistant, Siri and Alexa.
And what’s more, voice to text tools not only work well for English, but also for German. I have tested two services with which you can convert your own audio and video files into text within a few minutes.
Speechnotes Files
Speechnotes Files uses the voice to text engines from Google. To convert an audio file to text, Speechnotes Files takes about as much time as the file itself takes. Speechnotes Files promises, if the recording is excellent, an accuracy of over 95%. What I noticed in a test:
- Speechnotes Files also sets punctuation marks.
- Speechnotes Files recognizes names of known persons to 99%.
- In general I think the accuracy is very good, although Speechnotes Files is a bit worse in direct comparison with AmberScript (see below).
I would speak of an accuracy of about 90%. If a speaker swallows the endings, it would be desirable that the machine recognizes this and corrects the endings. Even words that are written in upper or lower case depending on the type of word, I often have to correct in manual post-processing.
Advantages of Speechnotes Files
- Relatively fast processing of the files (in the test the processing took only about half as long as the file itself).
- You can download the transcript as text file or subtitle file with timestamps.
- Privacy: Speechnotes promises that nobody else will have access to the files and that they will be removed as soon as the transcription is finished.
How much does it cost to use Speechnotes files?
- Each audio minute costs 0.10 USD.
- However, only predefined audio minutes can be purchased: 45 minutes, 120 minutes, 10 hours or 100 hours.
AmberScript
AmberScript is a start-up company from Amsterdam and Berlin. The voice to speech AI used is a proprietary development. According to the makers, because “there was nothing better on the market”. I uploaded the same file that I also transcribed with Speechnotes Files, and found out:
- AmberScript doesn’t like commas.
- AmberScript also recognizes names and important political terms.
- The accuracy of AmberScript on an excellent quality audio file is impressive.
- For lower quality files, AmberScript completely omits less intelligible passages (sometimes up to 30-40 seconds), so it doesn’t even try to transcribe at least the intelligible words.
Advantages of AmberScript
- Extremely fast processing of files. An audio file of 30 minutes duration was transcribed within three minutes.
- The transcriptions can be exported as Word, JSON, SRT, VTT, EBU-STL or plain text files.
- When exporting, there are various options to include or exclude, for example, time stamps or speaker changes.
- In addition, AmberScript provides its own editor with which the transcript can be edited manually.
How much does it cost to use AmberScript?
- There are different price plans for AmberScript: You can purchase audio minutes per hour, whereby one audio hour costs 10 dollars. One audio minute costs about 16 cents.
- Monthly plans have either three hours for 25 dollars or five hours for 40. A three-hour subscription costs about 14 cents per audio minute, a five-hour subscription about 13 cents per audio minute.
Conclusion
AmberScript convinces me a bit more about accuracy than SpeechNotes Files. The missing commas are quickly set, but with the editor you have a tool at your disposal that allows you many options in the postprocessing.
Absolute accuracy in automatic transcriptions is still not possible so that a manual post-editing is necessary in any case. Tools like AmberScript and Speechnotes Files make the work much easier so that it is easier and cheaper today to create text alternatives for audio and video.
Do you lack the time and experience for manual post-editing? I would be happy to support you in creating barrier-free content. Schedule a free consultation.
PS: Do you want free minutes for you and me? With the following affiliate links you can test AmberScript and Speechnotes:
Please also check gglot.com for automatic transcription. It is a startup from New York launched in 2020 during the global pandemic.
Hi Alex,
Thanks for your comment. Your solution looks promising, going to check it out to add it to my list of providers.