Using technology at our disposal, we can create accompanying transcription of any audio now.

[pon00050]

Hello.

There is an abundance of audio materials that one may be interested in.

As language learners, it's nicer to have accompanying transcriptions.

 

I have been using Auphonic for some time now but I found another automated transcription service yet that I want to bring your attention to.

 

https://sonix.ai/

 

To use Auphonic. you had to set up the Speech Recognition Integration service of your choice.

Sonix doesn't ask you to do that.

 

Sonix has far more features than Auphonic, including the feature to quickly and automatically create transcriptions and even subtitles if you need to.

If you already have transcripts, then you can also upload it to Sonix and you can get a subtitle file.

Check out there features here.

 

https://sonix.ai/features

 

 

So technically speaking, if there is a podcast series or anything that has audio component to it, you can easily create accompanying transcription using the two services I identified above. Both services are essentially paid. Maybe if you grab a friend or two who's also studying Chinese and wants to use the same resource as you are, then you all can use this service to create transcriptions/subtitles.

 

Happy learning!

[matteo]

Thanks for the suggestion, I use 有道云笔记 http://notesandbox.youdao.com/ on the phone to get transcripts of podcasts. It's free and surprisingly accurate.

[Insectosaurus]

Can it also provide timed subtitles for podcasts? Seems like it can, I'll probably try it out. Thanks for the recommendation.

[pon00050]

  

On 9/12/2021 at 10:22 PM, matteo said:

Thanks for the suggestion, I use 有道云笔记 http://notesandbox.youdao.com/ on the phone to get transcripts of podcasts. It's free and surprisingly accurate.

 

Thank you for sharing this. I didn't know about this. I don't feel comfortable downloading it though. It's nice to see that there is another alternative that is free!

 

 

On 9/13/2021 at 4:27 AM, Insectosaurus said:

Can it also provide timed subtitles for podcasts?

Yes! 

 

When you upload the audio, the service should automatically give you transcripts along with time-stamps.

 

On 9/12/2021 at 8:35 PM, pon00050 said:

If you already have transcripts, then you can also upload it to Sonix and you can get a subtitle file.

 

With this, I meant to say that you can upload transcripts and Sonix will align your transcripts and time-stamps.

https://sonix.ai/resources/transcription-realignment-realign-timecodes/

This also costs money.

But, yes it can be done.

 

 

So, for example, I hired someone to write a Python script for me. 

This scripts receives a subtitle file and audio file and outputs

a csv file, which contains two columns, one for time-stamps and the other for Chinese sentences, and

multiple audio files that are split from the original audio file according to the time-stamps provided in the subtitle file.

And then I would go through the text once using Chinese Text Analyser by Imron.

That allows me to identify the parts from that particular text that I need to learn from.

I imagine that I will be able to quickly identify these parts the more I use Chinese Text Analyser.

That is my study work flow for now.

[mikelove]

 

On 9/13/2021 at 8:27 AM, pon00050 said:

With this, I meant to say that you can upload transcripts and Sonix will align your transcripts and time-stamps.

 

This is neat, I'm going to have to experiment with it for adding timecodes to reader audio.

 

EDIT: tried it, didn't work at all - seemed to be trying to match up paragraphs based on silences without much regard for the content. (I could well imagine it working better for other types of audio, just didn't work at all for graded reader recordings)

[pon00050]

On 9/13/2021 at 10:57 AM, mikelove said:

This is neat, I'm going to have to experiment with it for adding timecodes to reader audio.

 

EDIT: tried it, didn't work at all - seemed to be trying to match up paragraphs based on silences without much regard for the content. (I could well imagine it working better for other types of audio, just didn't work at all for graded reader recordings)

 

I am sorry to hear that didn't work for you.

 

If you still have some patience for this, you could reach out to them and ask how come it's not working as it should.

 

I have been using Pleco for several years now and thank you all for the great work!

[malazann]

looks cool! There's a 30min free trial and am going to try it out. If this works it'll be a life saver!! The way some of you on this forum use technology is bloody amazing. Scripts, text analyzers, pitch recording. All very cool 

[glu]

On 9/13/2021 at 4:57 PM, mikelove said:

EDIT: tried it, didn't work at all - seemed to be trying to match up paragraphs based on silences without much regard for the content. (I could well imagine it working better for other types of audio, just didn't work at all for graded reader recordings)

 

I recently developed essentially the same thing for a friend who teaches Russian at a German university. She records prose in her own voice. I create a transcription of the audio using either Google's or Microsoft's speech API, and use a diff algorithm to match the transcription (which inevitably contains the odd recognition error) to the original text. This gives me word-level timestamps, which I can use to sync audio playback with the text.

 

Here is a sample page: https://prose.zydeo.net/player.html?ep=APT_BKR_1

 

 

[mikelove]

Thanks! I figure we can probably try to rig something like this up ourselves at some point - we have access to several different APIs for it.