AI Audio Upscaler - Low quality audio sounds high quality

It would be nice to have an audio upscaler AI model within VEAI and or even a standalone audio upscaler app for music and old recordings.

Sometimes old videos have low quality audio or the original audio wasnt captured correctly or formatted correctly and sounds off. This would fix that.

I’m just chiming in to say I would absolutely love this

3 Likes

Well, most DAW’s (Digital Audio Workstations) will allow you to “upscale” audio, by placing it into a larger container, like 32 bit float point. If you mean cleaning up bad audio or even reconstituting peaks, you can look into something like iZotope RX that specializes in that. But usually the process of audio restoration involves more than one feature and manual corrections, so I’m not sure to what extent could it be automated at this point.

If you have low quality audio in an old video, its less about “upscaling” it, because that is the easy part, the harder part is audio restoration and clean up. If you are in need of that, and you are waiting for Topaz to maybe implement something, you can look into iZotope RX, they specialize in that sort of thing and they are pretty much industry standard.

2 Likes

I guess, I was referring more to audio restoration.

Thanks for telling me about iZotope RX I think that can come in handy.

1 Like

Keep in mind this is a video enhancer not an audio enhancer. One thing I have noticed which is annoying, the audio that gets added to the upscaled video is most certainly worse quality than the original.

2 Likes

I am an avid user of RX-Advanced 9 (along with Diamond Cut Forensics and Adobe Audition). Me and my forensic colleagues are not very price sensitive, and we would all be active buyers of AI tools for audio super-resolution, and for the “cocktail problem”. The software must not upload case files to the cloud to be court compliant.

1 Like

Indeed. Some user-oriented media players, like PowerDVD, perform a remarkable audio-upscaling (aka, audio improvement), but I don’t feel VEAI should try and venture into that territory at all. For one, it requires extremely professional software, and VEAI can’t even copy audio right, it seems.

Dunno about others, but I always demux my movies to begin with, so I can process the video first, then mux the audio back in later.

With all the outstanding 3.x issues, I say let’s concentrate on Making VEAI 3 Great Again, and not get distracted with other things.

The topic has been on my mind again today.

I have many old drum and bass podcasts (mp3) from PFRadio which was taken off the net in 2015 and contains many records that are partly no longer available because the label no longer exists or the artist has died.

This is called Audio Super Resolution and there are a few githubs for it, but not a single software as far as I know.

The most important thing would be to take out the compression, you can practically hear the blocking you can see in pictures.

1 Like

It would be great to be able to restore compressed music or audio tracks to old movies from the 80s-90s. Especially from bad DVDs or video cassettes.

1 Like

This is what I use. It’s limited to two channels and pretty complicated to get setup, but I really like the results. Of course, I have not used it on many low quality mp3s. Mostly I use it to declip CD audio, but also the autoEQ is pretty good at helping oldie-moldie sounding recording to sound less so.

I agree with the statement that Topaz should stay out of audio, unless they tack it on for free until it’s actually something worth buying.

I’m dying for an audio enhancer with AI. Can you do it? Please if poss have one knocked up by next Thursday.

1 Like

I agree, especially because early versions of video a. i. wrecked my audio layers to ultra low 96 kbit/s. -.-

1 Like

I;ve learned to always keep the original from being bitten in the past.

I would like to add that ffmpeg does not have a great AC3 (Dolby digital) encoder and no DST encoder at all. If Topaz does do anything with audio enhancement, those two formats are crucial. My reasoning being that all DVDs and Blu-rays are in those formats.

If you’re into DTS Master Audio, you’d be better off with flac, if your equipment will play it. So I’m not saying that needs to be included.

1 Like

Would love this too.
THere are some applications that slightly clean up the sound.

1 Like

The Topaz engineers have taken this AI utility to amazing heights. Are there any plans for a future program that will do the same for Audio Refinement? That would be a blessing to those of us that are involved with Video production.
Dennis Freeman

1 Like

I have a cartoon that has a bunch of film grain on it. I feel like cartoons don’t need or benefit from having it, so I tried Artemis Medium Quality on it and it takes out the grain really well. I looked in a bunch of random sections and am not seeing any artifacts. It really looks like it’s just taking the grain out.

That got me thinking about audio. If that cartoon were to be audio, it would have very noticeable white noise like a cassette tape or a vinyl record. It’s steady and you tend to mentally tune it out, but it doesn’t have to be there. It shouldn’t be hard to make a model that can remove white noise from audio just as well as Artemis removes film grain from cartoons.

The biggest issue to this would be getting the audio back into Dolby and DTS formats. That alone would make the cost go up substantially.

they are already working tools to do that in the music world for this.
Steinberg Spectralayers, Izotope RX, Waves Z-noise, etc … so many…

1 Like

Oh I have StereoTool. It’s amazing once you figure out how to use it. Its declipper works much better and more automatically than Izotope RX—At least the trial version I tried of RX. It’s not made with movie audio in mind though.
What Topaz Labs could do is make something just for movie audio. It would need to be able to have Dolby/DTS in and same quality Dolby/DTS out. Some of the models I can think of would be 2 channel to 5.1, and noise reduction/removal.