iZotope RX9

[Guy Rowland]

...coming soon. But no further details as yet.

Grace period starts now for any purchases of RX8, you get the free equivalent version of RX9, RX9 Advanced or Post Production Suite 6 on release. Upgrade pricing from the previous version before the grace period looks like it is $99 for regular, $299 for Advanced or PPS6.

These threads - https://thesoundboard.net/viewtopic.php?p=65895#p65895 and https://thesoundboard.net/viewtopic.php ... 68&p=67205 have some audio examples of the miracles that can be performed in RX8, looking forward to seeing what 9 has in store. I was initially reticent to upgrade to 8 but it's been absolutely brilliant.

[lofi]

First update under the Native Instruments umbrella.

Will it be more music oriented fixes or will they AI-try to reach the excellence that is Cedar?

/Anders

[Guy Rowland]

First update under the Native Instruments umbrella.

Will it be more music oriented fixes or will they AI-try to reach the excellence that is Cedar?

/Anders

Actually I think RX is considerably better than Cedar (especially their newer D units where they couldn't get the same chips as the model before, and sound bloody awful IMO). BUT the really good stuff in RX is not real time so it's not a fair fight, granted.

I did only a few weeks ago request a feature that obviously can't be in this version, unless they'd read my mind good and early. A Clarity control, mostly for post. For really terrible quality buried lav mics, I often manually have to go through boosting the gain for all the Sss, Shss, Fffs, so all the other stuff isn't also boosted with horrible side-effects you'd get on poor quality material if you used regular EQ. I think with their dialogue detection algorithms it could be possible.

In fact such a control could be musically interesting too, so you could brighten or tame those very specific voice components without affecting the tonal balance of anything else.

[Arcana]

I only have RX 6 elements and the offer of upgrading to RX8 with the further free RX9 upgrade coming very soon, is a bit tempting.
My price is £97 via AudioDeluxe, but with Black Friday around the corner, it's tempting to wait and perhaps even get MPS 4 for cheap.
How does that upgrade price sound? It's hard to tell with Izotope as they have new offers every other week.

It seems like the best tool for de-noising, de-clicking and other types of noise removal.
Probably quite essential if you record vocals and guitar reguarly.

I'm not sure it has the best music rebalance feature out there - I think RipX deep audio perhaps wins this category. It's not a feature I'd use much though, apart from removing vocals from a commercial track in order to just hear the instrumental though.
Actually Cubase's own Spectral layers is actually quite good as this task, even if it only split the songs into vocals and music - ie. 2 stems.

[Guy Rowland]

£97 sounds good to me - I don't think an upgrade has ever been below that (could be wrong). The guitar de-noise is really helpful for controlling the squeaks at a touch - lovely.

I haven't played with RipX. RX8's rebalance is really good though (considerably improved from 7), there's some examples in the threads I posted in the OP. Spectralyers is pretty good too, but as you say only Vocals.

[lofi]

First update under the Native Instruments umbrella.

Will it be more music oriented fixes or will they AI-try to reach the excellence that is Cedar?

/Anders

Actually I think RX is considerably better than Cedar (especially their newer D units where they couldn't get the same chips as the model before, and sound bloody awful IMO). BUT the really good stuff in RX is not real time so it's not a fair fight, granted.

I did only a few weeks ago request a feature that obviously can't be in this version, unless they'd read my mind good and early. A Clarity control, mostly for post. For really terrible quality buried lav mics, I often manually have to go through boosting the gain for all the Sss, Shss, Fffs, so all the other stuff isn't also boosted with horrible side-effects you'd get on poor quality material if you used regular EQ. I think with their dialogue detection algorithms it could be possible.

In fact such a control could be musically interesting too, so you could brighten or tame those very specific voice components without affecting the tonal balance of anything else.

I'll have to disagree with the Cedar part, they are miles ahead when it comes to instant results.
Time is money and all that... But they don't really play in the same league, so sorry that I brought it up.
And don't get me wrong I use RX8 everyday, I wish they'd stopp adding stuff and dedicating all their resources to better realtime noise reduction.
Dialogue de-noise could be the best thing ever. Now it's what producers/creators expect us to use and it so damn flat sounding.
Last few series I worked on had no de-noise on them. I was more thorough with EQ and offline processing. That gave a better result, but obviously took longer to mix. Didn't get extra paid for it but I'm so sick of the RX "sound" so I did it anyways.
...
Sounds like an interesting feature request and probably what would give a better dialogue de-noise result.
Hope they used AI to read your mind so that we'll get Clarity to our mixes. I know I need it

/Anders

[Guy Rowland]

I should have said - agreed dialogue de-noise is no good, I never use it. I mentioned it wasn't a fair fight though, and that's why - I was comparing offline best quality RX with real time Cedars.

The older model Cedars are really good, but apparently they could no longer source some of the chips so had to redesign internally, adding the D suffix. These to both our ears on a show last week sounded horrible, took all the life out of it and pumped. A bit like Dialogue de-noise, funnily enough! I've yet to hear anything better than offline best quality RX though, using a manual profile. Any improvements to real time would be welcome of course.

[wst3]

it has been a long (LONG) time since I had the opportunity to use CEDAR hardware. At the time it was borderline magic, nothing came close.

Now RX, Accusonus, and a couple others get close, to my ears iZotope gets the closest, but again I haven't used the current crop of hardware. AND, for my purposes I am perfectly happy with saving a couple bucks by using RX. Currently I am using RX7 because, well because I've had no call for it in about a year and a half. When live theatre really opens up again that might change.

As good as all the iZotope tools are, I do have a bone or two to pick with them, not the least of which is their bizarre loyalty rewards program. I have a bunch of offers, some of which are not at all applicable, some of which make little sense, and the rest, well, the rest are not attractive deals. I wish they'd return to simple, easy to understand, and really tempting loyalty rewards.

[Guy Rowland]

It's out and I am potentially ecstatic, though the headline is - this is a Post update, not a music one.

It's sorta light on the number new features, but two of the three are potentially huge for me, algorithm improvements to Dialogue Isolate, Ambience Match and hum. Dialogue Isolate is pretty miraculous as it is, but the demo on the website - https://www.izotope.com/en/products/rx.html - has me salivating with the clarity. It's hit and miss at the moment and often compromises quality. Ambience Match has been mostly useless to me until now because it has a frozen effect on the background atmos, it really only works on very static sounds like aircon (which you can usually easily get rid of anyway). This now claims to work on complex material - I'll be very keen to try it out, that could make me weep for joy if it works as well as it does in the demo on a wide variety of material. De-hum now works with all kinds of hum not just standard mains-induced kinds which might be handy, but there's usually other tools in RX that can do this anyway. (Tangent - working on a show right now that had horrible hum over everything - frequencies turned out to be exactly 100, 200, 250 and 300hz. No clue what caused it, but Neutron EQ makes it vanish completely just using very high Q values, that's inserted in the mic group so need for RX there at all!)

There's a couple of other feature updates like a history list, restore selection and ARA but for Logic users only.

Compelling update for Post, one to skip for Music. Upgrade price is half the usual - $149 for me for Advanced. Not sure about Post Production Pack 6 upgrade yet, that's not up on the website as I type.

[soundbylaura]

Not a lot of new stuff for a full-number update, if I'm honest.

(Advanced 8 to Advanced 9 - $125.16 at JRRShop with code GROUP, but it's early still, not all the resellers have it up).

[Guy Rowland]

Ooh thanks for the JRR tip Laura. Tax cancels out the discount here, but it would probably be added on at the iZotope shop, so bought from them anyway.

(buying / install issues - skip if not relevant) You have to go through a mystery link, input your "serial number" which isn't a serial number, and then it generates your real serial number. That went fine into the Product Portal, but weirdly RX9 showed up without any install option. I downloaded the demo version from the product page instead, and that seems to work, picking up the real license on the iLok (once I'd copied it from my account... the installation suggested it would dump it right in the iLok but it didn't).

It's late here - way past my bedtime - so I've only given it a few demanding tests on 2 of my current shows that need plenty of NR. And I have to say, first results are.... terrible. In every case the legacy Dialogue Isolate sounded either about the same or - usually - far more natural, and the new algorithm took about 5x the amount of time to do it. I tried Ambience Match on a really tough one, a water jet over the dialogue. It didn't know what the hell was going on, essentially useless.

I'll be using it in anger over the coming week and that will determine whether or not it really is a bust - I always find some things work better or some material than others. But one thing's for sure - it doesn't work on all material like it does on the website demo.

[Guy Rowland]

Just had a thought - I may not be able to post the clip, but I can share the images.

Here's an offending original clip - presenter on a water jet thing. See how clear the trace of the dialogue is.



Here's the settings (RX9 shown here but same for 8)



Here's the results in RX8 (legacy):



Pretty crude, but it's mostly got the basic idea. Here, however is result in RX9:



You don't really need to hear that to know how bad it sounds. And remember - this is Best quality, takes 5x the time to render, and is the headline feature for RX9. (I tried it at the lower quality setting - no difference).

[lofi]

Looks awful.

I have the opposite experience with RX9.
Did a much better and faster job than RX8.
Although I have to say I’ve never used that amount of reduction in dialogue Isolate.
I usually run multiple passes with lighter/lesser values.

Dialogue denoise is as bad as 8.
I’m a fan of the McDSP denoiser but that’s off topic.

I paid 130usd for the postproduction suite.
I’d say it’s worth it.

Really liked the new de-hum plugin.
Was able to get rid of some noisy light within seconds.

/Anders

[Guy Rowland]

Thanks Anders - faster does surprise me, others are reporting it to be glacially slow in Best mode (Good mode is about the same as RX8 for me, speed-wise)

It's worth doing a full aggressive pass to hear exactly what it is doing. even if it's only a test - less aggressive fools the ear more easily of course. Regardless, it has to be that some material is responding better than others - you clearly are getting good results. I actually imported their demo clip up on their website which is set to 100% and it sounds just as they present, so there is improvement on some material. Oddly that demo material is much more complex than my example, but it makes a better job of it.

[lofi]

Thanks Anders - faster does surprise me, others are reporting it to be glacially slow in Best mode (Good mode is about the same as RX8 for me, speed-wise)

It's worth doing a full aggressive pass to hear exactly what it is doing. even if it's only a test - less aggressive fools the ear more easily of course. Regardless, it has to be that some material is responding better than others - you clearly are getting good results. I actually imported their demo clip up on their website which is set to 100% and it sounds just as they present, so there is improvement on some material. Oddly that demo material is much more complex than my example, but it makes a better job of it.

I'm sorry if I was misleading.
I haven't really used the Isolate function on RX9
But the others de-click, -hum, -clip etc was faster on RX9.
Not in the studio but could AB RX8/9 if anyone wants numbers
I'm on a modest Mac nothing fancy.

My guess would be that the AI was trained on the demo material.
We Swedes use the front of our mouth when speaking, Americans use more of the back of the mouth.
Some posh Britts hardly move their mouth all when speaking etc...

This is of course only speculation, but I believe that the AI has been trained more on American voices than on European.
That might be why non-AI noise reduction works better with the material I work with.

Anyhow, keep reporting your findings.

Best,
Anders

[Guy Rowland]

Ah gotya, thanks Anders. I haven't tried any of the non-changed plugins yet, assumed they'd all be the same - glad they're quicker!

BTW, "good" mode in Dialogue Isolate works at the speed of the legacy algorithm, and thus far no-one has been able to detect a quality difference (!).