Apple transcribes podcasts in writing and it’s far from stupid


The Apple Podcasts application will receive a new feature with iOS 17.4: written transcription of podcasts. French is supported.

At first glance, one might find it curious to transcribe a podcast in writing. Why read a conversation that was designed to be listened to? However, this option arrives in Apple Podcasts, with the iOS 17.4 update (currently only available in beta).

In reality, this idea is very practical. It has at least two advantages: the first concerns accessibility, since the podcasts broadcast by the service of the Cupertino company can also be understood by deaf and hard of hearing people, by giving them access to the automatically generated text. Transcription takes just a few seconds, according to Apple.

The second makes it easier for listeners to use a podcast episode. Do you want to search for a specific keyword? Podcasts allows it and will tell you if it finds it or not. If so, the application gives you the possibility to go right to the place in the podcast where it is spoken. Journalists, for example, will appreciate it. People looking for a specific time too.

Apple podcasts transcript
An overview of the functionality. // Source: Apple

Apple Podcasts manages the transcript in French

The written transcription of podcasts is based on a well-known process called automatic speech recognition. We find this technique when a person asks a question to a virtual assistant like Siri. The tool captures speech, identifies the words, understands them, processes the request and responds with a synthetic voice.

This transcription option is transparent for podcast authors. Apple specifies, however, that the audio quality of the file must be of good quality for the tool to work. If necessary, it is possible to provide the text file yourself, but also to make corrections if passages are erroneous.

According to Apple, this automatic speech recognition obviously manages to distinguish what comes from the speakers and what comes from possible background sound, if ever a song is picked up. “ Music lyrics are also not displayed in transcriptions ”, it is thus indicated.

The Podcasts application will be updated globally (170 countries and regions are eligible, according to Apple), but language support is much tighter. Currently, four languages ​​are supported: English, Spanish, German and French. Enough to already cover a large number of speakers around the world. It is also a first step towards Apple’s first generative AI, expected in June.

For further

Excerpt from the 1st episode of the podcast.  // Source: NumeramaExcerpt from the 1st episode of the podcast.  // Source: Numerama


Do you want to know everything about the mobility of tomorrow, from electric cars to e-bikes? Subscribe now to our Watt Else newsletter!



Source link -100