Adobe Podcast Enhance Speech languages

[Jeroen32141884k4dp]

Does the Enhance Speech tool in Adobe Podcast only works with an English speaking voice? Or are more languages supported? 

[Nancy OShea]

[Moderator moved from Using the Community forums to Adobe Express.]

More features are available with a Premium Adobe Express Plan but I don't think Podcast Voice goes much beyond English yet.

https://podcast.adobe.com/features

[Faz Audiovisual]

Just from my own user perspective and experience: I've used this a lot mainly with Spanish (besides English, German and French). It has been very useful for many recordings. But I started to notice that when the original recording is weak and/or not clear enough, the audio artifacts tend to sound more like American English babble. So I thought/understood that it seems like this application is not a sound cleaner, but a speech reconstructor. That's why it may need to be trained for specific languages. I insist, I am saying all this just from what I have noticed so far as a user.

[eliezerr34934272]

I tried witha Spanish audio track, the result was trash, but that was many months ago when it was initially released. Crumple Pop is the best audio enhancer I've tried, so far.

[Rocio34160521nw9k]

yes, in spanish is really bad, i hope the v2 (release in october) fix that

[Harmony J · Accepted solution]

Hi @Jeroen32141884k4dp , Harmony here from the Adobe Podcast team 👋

Enhance Speech is language agnostic and does not generate or translate across different languages. The results often depend on the speaker's audibility and the amount of background noise. It’s important to note that clearer and more polished inputs (speaker pronunciation and enunciation) will produce better outputs. This tool enhances audio by filtering out noise and artifacts, adjusting pitch and volume levels, and normalizing the audio.

If you haven't already, you can sign up for early access to Enhance Speech v2 here: https://podcast.adobe.com/enhance-speech-v2

Don't hesitate to reach out with any additional questions or feedback!

[Faz Audiovisual]

I think that you are right in the sense that the model was not developed specifically for English. But it is likely that the data used in the training had a bias towards English. To me, that was evident. For example, whenever I had a voice in the background besides the main speaker, that is, a subtle voice behind during a silence, the model tried to rebuild speech that clearly resembles American English gibberish. I also think (or hope) that this new v2 will improve in all aspects. Anyway, this has been a GREAT and very useful tool!

[�ryan_8789]

No

[sheetal_8773]

Aaj ka din bhut acha hai 

[Asher30859326113b]

 

 

[John35759503l6ib]

V1 was useless with Turkish - turned it into Klingon... Need to try v2...