Hellloooo! So good to be back :) Did you try latest Cedar Audio plugin yourself? What do you think about it? And about its price tag? Lemme know down below!
I hear more like spectral swishing/swimming type artifacts on the actual dialogue for Waves and Acon. The actual dialogue quality sounded more natural with Cedar to my ears. In terms of background artifacting, I think Cedar is biasing natural sounding dialogue over gating artifacts. Goyo has a similar behavior. Acon is is removing a lot of the noise at the cost of large chunks of the dialogue's frequency spectrum missing where the NR is removing noise that is masking the natural dialogue, which results in that swimming artifact. Something like rxRevive would help because it resynthesises a lot of the dialogue spectrum that's being masked by the noise. @2:40 "my new". Heavy artifacting on new with Acon as an example.
You commented about artefacts but you are using the maximum background reduction. Have you tried other settings? Also try less background reduction with levels of voice level increased.
[AN ADDITIONAL WAY]: If your recording setup is able, try placing the C-SUITE C-VOX NOISE & AMBIENCE REDUCTION plugin from UAD (which is also a noise suppression technology from CEDAR) at the beginning of your live recordings and then try to utilize the CEDAR AUDIO VOICEX plugin in "post" to clean-up any remaining noise. NOTE: This may be a cheaper way to get high quality noise reduction by utilizing the CEDAR technology and without using the more expensive CEDAR AUDIO STUDIO COMPLETE bundle.
Have you tried running multiple instances on these types of plugins to reduce artifacts and keep the vocals/ dialog more in tact?b also what a lot of these types of videos fail to do It is test for more real world situations that home producers might encounter. Such as not having a well-treated room and having small amounts of background noise. I'm wondering if cedar Audio might actually be preserving more of the natural voice... I'm not sure about that. I should do some tests of my own.
I did similar tests and disagree. Cedar VoicEX sounds more natural to my ears, keeping the original tone of the voice, while the others alter the tone of the voice to sound more like a Zoom call. Some other people in the comments also pointed that out. Also, I almost never reduce noise to 100%, as the results usually sound way too processed. On lighter settings like 50% reduction, Cedar VoicEX was always performing better than Supertone Clear (formerly Goyo), Accentize, rxRevive and Hush Pro on my recordings which had a lot of crosstalk. It was the only plugin I could reliably use during batch processing on many hours of material. All other plugins showed unacceptable artefacts when I did some listening checks. Still think the price is steep and so far I didn't buy it, just used the trial for one project.
Oh wow - have they done any updates? I just got the free try of it and it seems good - but, I'm really liking Accentize products at the moment - over $1000 and it only has two knobs, no EQ splitting... yeah just not sure.
@@FoliaSound thanks for the reply! Yeah, Clarity is fantastic, along with Accentize, and Crumplepop. I also have the full RX advance suite, but mostly use that for spectral dialog edits since their denoise is a bit behind the curve now. I've been looking at Acon for a while, good to hear that they are solid - very inexpensive, so I'll pick up the suite.
Seriously? There's absolutely nothing about it in the manual whatsoever. Moreover, this is what the manual says: "(VoicEX) incorporates some unique algorithmic features that take it beyond standard DNN-based noise reduction techniques and, for a wide range of material, is able to separate voices (primarily but not limited to speech) from other sounds".
I wonder who engineered that plugin. The original Cedar guys were so clever but 30+ years on their original algorithms haven't been updated for modern higher resolutions and this new stuff seems mediocre. It's like they're a sales business now and nothing more. Maybe they're just surviving off their legacy and customer base? I dunno. Anyway now is the time for someone to step up and take over their market
I'm perplexed by your verdict. My impression was entirely different. From everything I could hear in your 100% examples, both the Waves and the Acon plugin affect the sound of the voice a lot more negatively than Voice-Ex. They produce artefacts in both the voice and the ambience. In all of your examples, Cedar's Voice-Ex left the voice a lot more natural and unaffected than the others. In your 2nd example, it would have been very helpful to hear the noisy original. It would have also been helpful to listen to the plugins while you turn them from 0 to 100, not only at 100%. It depends on what you want from a noise reduction plugin, either low price or quality of voice. When it comes to price, Acon and Waves are the winners. If you want naturalness of voice, VoiceEx is the clear winner. If you're familiar with Pareto's principle, it's clear, why the Cedar plugin needs so many more resources: the extra voice quality requires a lot more computing power. The Waves and Acon plugin are satisfied with 80%, while Voice-Ex tries to reach at least over 90%. That requires exponentially more computing power.
Hey, thanks for your comment! The voice might seem more natural with Cedar, but artifacts and inconsistencies in the denoising process are unacceptable as for me. You can achieve same voice quality and way better background removal job, just decreasing processing amount on both Acon and Waves. As for optimization, I'm afraid you're quite wrong here: all machine learning output code works on a principle of a black box - you know what you put in, but you don't know what you take out and how it happens. Machine learning process excludes any human, on-purpose coding and proceeds absolutely independently from human intervention per se. You can only set up starting conditions and material used and then, after the learning is done - try to optimize the code, which is really difficult, as the code is hardly understood by any human programmer. Thus the fact that Cedar eats up so much resources, doesn't come from the fact of its quality, but rather from the fact of very poor or no code optimization at all. It indicates very little work put into it and absolutely no serious testing conceived prior to the release. Having any AI plugin that makes such demands on a modern CPU has nothing to do with quality, it only displays poorly conducted ML process and a complete blunder when it comes to pre-release testing. I'm really sorry to say all this, but again - thanks a lot for your nice, in-depht and cultured comment, this is how people should discuss things online! P. S. I already saw some people buying Cedar denoiser, being led by the brand, not testing out its competitors and then defending the money they spent. It's a sorry thing to watch for me, but this one is a serious loss for Cedar and I believe it substantially shook the foundations, which this company has been built on. Have a Merry Christmas and a Happy New Year!
