This app is built using a pretty complicated algorithm for noise to speech. With the processing power of todays mobile devices, it is therefore not possible to make the speech clearer or analyzing it further in order to produce a textual context. The app would then be so slow that it would be unusable for most users and any longer being able to play sound in real-time. It would also require us to do more interpretation in the app and by this introducing more false positives; something that we don't want.
There are already a lot of apps in this category in all of the app stores, and we don't want to become like them: "random text machines".
Update 2016-05-30: Updated algorithm for speech to text
- Turn up the volume
- Use an external speaker (either Bluetooth or with cable)
- Wait until the app is fully calibrated. The calibration bar is shown in the bottom of the TV screen.
- Have patience! Some users have reported that they get very clear speech, while others experience that they don't understand anything at all or that they never hear anything. Time and Context, seems to be what differs most between these two categories of users. Please, give the app a chance and try it out for more a few minutes. Instead run it several times (15m+) during a longer period of time. If you still don't get any or clearer speech, please contact our support.
In version 2.0.0+ of the app, the upgraded text translation shows a percentage next to the word or phrase being identified. The number indicates how likely it is that the analyser has identified the correct word. In general, everything above 50% is considered having a high probability.Update 2015-02-02: Approximative algorithm for speech to text
Version 1.1.7+ of the app has voice to text translation based on an approximate algorithm, which means that the app will do its best to understand what is being said and show it on the TV screen. This means that sometimes only a part of the speech is shown. The translation algorithm is still in an experimental beta phase, and will be updated.
Three dots "..." are used to show you the position of the identified part. If no dots are shown, the speech was completely translated.
- Beginning: Some characters...
- Middle: ...Some characters...
- End: ...Some characters
Please note that this is only and example and does not
reflect actual app content!
Speech: "Cows are green"
- Beginning: Cows...
- Middle: ...are...
- End: ...green
Last updated Tue, Sep 29 2015 4:01pm