Adobe Podcast AI Sounds Bad

DO NOT contact anyone on Telegram (or other service) that is pretending to be me and claiming you won something. I am not giving anything away. They are scammers trying to steal your money and data.

This is a very hilarious video. 😅

I LOVE the sound of podcast! Saved me from having to do some ADR on my film

It gets better the worse the source audio is.

I Think it's work just with men.. When i speak my voice become more masculine and so weird.. Is there is any other ai i can use?

Is there any other ai enhancer that could clean? if we as a species develop ai these are the kinds of problems that should be fixed.

This is so very helpful. Thank you!!

Have you heard Waves' Clarity reverb remover in action? I think it's an absolute god send. Though, I imagine anyone using it would still need to have some acoustic treatment in their space to get the most out of it.

I record speakers at my Toastmastets meeting each week. It's a big room with 30 people coughing, shuffling their feet etc. It is way too much rigamarole to lav people up, so I grab sound using a Tascam recorder on the speaker side of the tables. It's understandable, but does add a hollow feel to the speakers' voices. Lately I've been running the audio though this tool and then mixing it in with the live room. The output from this adds fullness back to the speakers voice, which when mixed with the room sounds natural.

Open question, feel like this might be a good place to ask. I'm looking to listen to old episodes of a podcast but they're really hard on the ears. is there something that can clean them up in real time while I'm listening?

Honestly this taught me that the iPhone mic is pretty much good enough for most usage (obviously not pro podcasts or anything tho) on its own

Um...sorry but this tech is INCREDIBLE. Most people are going to basically give you the first example or have a nice mic but not treated in the right room. This is magic for exactly the kinds of people who will use it.

great job

I can never get good results with this - perhaps because I am running audio through it that contains a voice with an Australian accent. It seemingly attempts to Americanise the voice and will change words such as "happy" to "fappy" which is not desirable haha.

The only one that I thought was better was the clipping (and barely at that). The AI processed audio just sounds to synthetic, and removes all the expression in the voice, kinda uncanny valley but with vocals.

Adobe Podcast AI sounds muffled to me, not robotic. It keeps your pitch, but not your dynamics, as well. Honestly, I’d use it for narration from my phone mic in a hotel or something, then run it through an EQ to fix the muffled sound.

Unfortunately, it's very hit or miss

thanks Kendall Roy for explaining this

I'm curious how it compares to RTX voice.

I actually find the raw iphone sounds better (given the recording situation is good)

I would use both the audios, the highs of the raw + the AI for the lows

I think it might be usable

It's pretty good at removing reverb'

Don't agree with the robotic comment. Everytime I've tried it, it sounded pretty great.

Must say I love it. A lot of the time I get audio send in which is recorded very badly. Very noisy for example. In this cases Adobe can fix things the expensive iZotope RX can not fix. It also levels, which is not always perfect. And what I don't get: although the EQ sounds warm and nice it lacks a little top-end. Which can be easily fixed, but it sound a little muffled.

some of them sound better. The background removal is decent

100% agree! I just tried this and you are absolutely right... It's too aggressive and my voice is unrecognizable! I recently did an important interview with a sports star and I nearly cried when I realized all I was left with was a normal DSLR sound. Is there ANYTHING better and simple to use that you could recommend? I'm really struggling here and not very tech savvy! 🤦‍♂️

I have to agree. It does amazing things, but it's still riddled with artifact that a professional would never tolerate. Still no substitute for a properly treated space.

It seems to create weird artifacts with "th-" and "st-" and "-cess" sounds even without extra challenges of background noise or the other sound obstacles you tried.

สะเหล่อ

The last few audios of my content are done with this tool. Sometimes it sounds weird in specific moments but very rarely. Other than that I think it really does a good job for me fixing my horrible mic.

very nice and informative video, just want to add that Adobe podcast ai makes stereo audio to mono instead which is kinda weird too hope they will fix that

Did you separate the MP3 file of this video and then run it in Adobe podcast AI and then merge it back with the video?
Adobe podcast AI only allows the audio.
Can you help me clear the video voice aswell.
Please recommend what I need to do.

I only use this because my headset microphone (Corsair gaming headset before you ask) tends to have a static (tsssssss) in the background. This program knocks that static out and even stops the crackling it has. It’s great

The phone call audio thounds kinds weird.

it did great in some situations but in others it was pretty bad, it did destroy the good recording which sucks. one day someone will get something like this working right.

Interesting to me that Adobe BAI is making choices. There is a consistent spectrum on each sample. Even though they sounded profoundly better, I’m fascinated that someone chooses to calibrate the AI at which point it’s not really AI.

I'm not tho thure if I like that prothething on that phone call

I think it works well for shotgun mics in untreated rooms, did ok with removing AC/ fan noises if needed. Wouldn’t touch it for anything else.

Don’t watch the phone call AI part with coffee. For the sake of all things holy… brace BRACE BRACE!!! Lol.

I've worked with this thing for quite some time.
In my projects, I have lot of archival recordings made on old hardware and sometimes bad takes happen.
In most case it DOES sound robotic. When trying to restore things things, that's bad.... But.....

Here's what I've found.
It's wise to get back to your DAW and sometimes mix both audios in certain ways. Sometimes 50/50 is good.
But much better is to select frequencies which have some problems and replace them with AI sound.
For example, we're fighting with reverb. But does reverberation rings through the entire spectrum of audio? Probably not....
it always has some dominant resonating frequencies. Like somewhere in the middle, maybe middle-bass(room resonating frequency). That's it. Usually there is no point in changing bass or treble frequency ranges. They can be left intact.
My point is - to experiment more with frequency ranges and mixing. Sometimes you can get pretty good result if you tinker with it long enough....

There is no one button solution and probably there's never will be one. Machine doesn't know what do you want.

Sounds amazing

I'm having bad results with Adobe Podcast. It's sounding really artificial and digital. And the artifacts are very annoying. However, I'm sure they will still improve things.

It worked well for me, I just want to create random tutorials on my channel. No one cares if my voice isn't perfect, as long as there isn't a dog barking in the background. That would be perfect for me.

What I find this useful for is recovering high frequencies from very lossy recordings - think restoring audio from the early digital days that may be recorded at 32Khz or lower. When you mix the "enhanced version" highpassed around the nyquist frequency of the original dialog, it does have some success at making the recording sound more modern and "better". I wouldn't use the AI processed versions on their own though, they do sound robotic.

I've had trouble getting it to work with female dialog well though - it's possible that the AI model is mostly being trained on male voices so it works with that better for the moment.

I work in the tech field in a small-medium sized company, so whenever people do recordings at home, the audio always sounds like they are using their mid 2000s phones to record. This application has saved me MANY times now having to edit these recordings that are almost unsaveable with my skill (since i'm not anywhere near an audio expert/engineer). If at any point the audio is way too bassy and processed, then i go in to adjust it and dial it down to a less robotic sound.

It's been very useful for cleaning up badly recorded tracks on podcasts I edit. The tendency to make everything sound the same was also quite useful when I lost a section of one track and had to replace it with audio recorded from an iPhone.

Try it on a track where someone is talking over other voices, like we had at a recent live event. It comes up with the same voice speaking nonsense syllables, which is fun