Enhance Speech v2 is here!

[Harmony J]

We’re thrilled to introduce Adobe Podcast Enhance Speech v2 to our community. Our team has been hard at work to bring you a tool that transforms your audio and video recordings into studio-quality sound with just a few clicks.

Here are the v2 highlights:

Natural clarity: Clear, natural dialogue without robotic tones—no matter the language or recording conditions. It handles everything from noisy crowds to poor acoustics and delivers studio-quality audio.

Strength slider: Powered by our Sound Lift AI model, the updated slider allows you to balance speech clarity with ambient noise. Want a bit of café ambiance in your podcast? Adjust the slider to find the perfect mix.

Noise removal: Background noise, reverb, and music—v2 removes it all, keeping your message sharp and clear.

Clarity boost: Struggling with quiet or distant audio? V2 improves clarity in soft recordings without amplifying unwanted noise.

Try out Enhance Speech v2 here:

podcast.adobe.com/enhance

We’d love to hear your feedback and suggestions. Your thoughts are super valuable as we continue to improve and innovate.

We’re excited to see how you use v2 to elevate your audio storytelling!

[Pavel Soloviev]

Adjusting the settings (moving the sliders) doesn't affect the recording at all. Am I correct in understanding that the sound should change when I adjust settings during playback? I don't hear any difference. It feels like the same file is playing. It's about both v1 and v2.

[Harmony J]

Hi @Pavel Soloviev you should definitely hear a difference when adjusting the strength sliders. Keep in mind that with v2, you'll have the ability to independently control speech enhancement and background noise. If you have a file that you can share, that will be helpful in troubleshooting. Feel free to send it to me at hjiroudek@adobe.com

[Johne_Leon0877]

hi, everyone
version V2 so called AI is bad. She change my voice and so most important are cutting and change letters in my speech. wth? also change sound volume as wave. GREAT ADOBE WHAT`S WRONG WITH YOU???

[Adrian_DeWitts5625]

Hi Harmony,

v2 has been impressive and flawless for our use case.

The 4 hours are limiting for us for the amount of speech we need to process. Is there a roadmap on v2 to run in Premiere or as a standalone plugin on our own machines? I imagine this probably requires a large GPU.

Thank you.

[Harmony J]

Hey @Adrian_DeWitts5625 

I'm so glad to hear v2 is working well for your use case! I understand that 4 hours per day is less than ideal for some use cases. As we scale and add new functionality, we will update our limits over time.  
 
You can also upload all your files together, and we will automatically start enhancing them the next day when the 4-hour limit resets.  Hoping that reduces some of the overhead of uploading files daily!

At the moment, Enhance Speech v1 in available in Premiere. The goal is to eventually make v2 available in Premiere, but I don't have a timeline on that at the moment. I'll certainly keep you and the community posted as we make progress on this. 

[Adrian_DeWitts5625]

Thank you, Harmony. Fingers crossed v2 in Premiere happens soon.

[Roy_Gordon4903]

The last day, I have been getting enhance fails messages v1 and v2.

1. Why do I keep receiving those failure messages (it's the first time I've received them)

In addition, what is the reason for using my credit if it is a failed enhancement?

Does anyone have experience with this behavior

[Harmony J]

Hi @Roy_Gordon4903 Thank you for reaching out! I'm not able to reproduce this issue – would you be able to share one of the files this is happening with? 

[Roy_Gordon4903]

  Just before uploading the file, I tried to enhance it again. It's going well so far!  

[kalpanax]

v2 doesn't seem to recognize the entire vocal correctly, leading to a lot of 'drop outs'--moments where the audio cuts back and forth between enhanced and un-enhanced audio. I like the warmer sound of v2's enhanced audio--I'd just like to be able to extend it to the whole track. Otherwise the end result is filled with tantalizing moments of how great it could be, but ultimately it is not usable.

v1 recognizes all of the vocal, but I like the warmer v2 sound. Are there any ways to help v2 recognize all of the vocal? I tried adding some light background noise to heighten the contrast and help v2 distinguish between voice and background, but it didn't work.

Also, could this feedback be addressed in potential updates / in a potential v3 someday? Should I submit a copy of this using the Contact Us link, or by posting here it might be relayed to the relevant team? Thank you 🙏