Hello everyone,

I’m one of those who are very critical about the Pianoteq sound but I’m probably not as good as you at tweaking the sound. I stick with Pianoteq nonetheless because I think that they use the currently most promising technology. I haven’t listened to your tweaked presets yet  (which I will probably do later) but I noticed something interesting playing around with the EQ. It seems like the metallic sound comes from the region somewhere between 1560 Hz and 2250 Hz. So my EQ settings are 1560 Hz, 0 dB; 1820 Hz, -13.0 dB; 2250 Hz, 0 dB. Something I also did was to slightly increase the bass frequencies below 400 Hz.

Even my mother who usually says that she can’t hear that Pianoteq sounds artificial said that she preferred my tweaked sound by far. I applied the EQ settings to the NY Steinway D Classical preset and the Kawai SK-EX Ryuyo preset and I think both sound better now but are certainly not perfect. I hope that this might help somehow.

Thanks. What’s your Q factor for those EQ points?

I’m one of those who are very critical about the Pianoteq sound but I’m probably not as good as you at tweaking the sound. I stick with Pianoteq nonetheless because I think that they use the currently most promising technology. I haven’t listened to your tweaked presets yet  (which I will probably do later) but I noticed something interesting playing around with the EQ. It seems like the metallic sound comes from the region somewhere between 1560 Hz and 2250 Hz. So my EQ settings are 1560 Hz, 0 dB; 1820 Hz, -13.0 dB; 2250 Hz, 0 dB. Something I also did was to slightly increase the bass frequencies below 400 Hz.

Even my mother who usually says that she can’t hear that Pianoteq sounds artificial said that she preferred my tweaked sound by far. I applied the EQ settings to the NY Steinway D Classical preset and the Kawai SK-EX Ryuyo preset and I think both sound better now but are certainly not perfect. I hope that this might help somehow.

daniel_r328 wrote:

dikrek wrote:

This is also for daniel_r328, replacing that reverb with the QRS, which is supposed to blend the sound better with the source.

This I can confirm - QRS sounds more organic to me

If compared to the More or Less Hammer, I totally agree.
If compared to the original one(s), I can barely hear a difference.... and I have to go musical phrase by musical phrase, alternating between the two or three options, often multiple times to hear that they are not exactly the same!!

Then 2 QRS versions have more or less hammer tone, too similar then?

At this point I do t know what else to tweak.

dikrek wrote:

This is also for daniel_r328, replacing that reverb with the QRS, which is supposed to blend the sound better with the source.

This I can confirm - QRS sounds more organic to me

If compared to the More or Less Hammer, I totally agree.
If compared to the original one(s), I can barely hear a difference.... and I have to go musical phrase by musical phrase, alternating between the two or three options, often multiple times to hear that they are not exactly the same!!

This is like going to the optometrist!

This exactly is!!!

Not sure what you did with these QRS versions (the only QRS I know is PNOmation). Whatever you did, you wonderfully removed the muffling, completely, so great job there. On the other hand, these QRS versions to me sound almost indistinguishable from my favorite which still is dv_5th_NY_Steinway_Jazz_modified.flac (exactly that optometrist feeling when they propose you 3 different lenses and they all seem identical to you, eh?)

I did try different speakers (don't have any better decent headphones), and while the signature across devices is different, the "problem" I hear is identical. Going back to vision, I think you are concentrating on fine details (say if the corners of the E are sharp 90 degrees angles or rounded) whereas I am concentrating on gross features, say if the E is slanted as in E. Oh well, we can actually even claim that our role are reversed (but I didn't want to say that you are doing "gross" stuff, I prefer saying that you are doing "fine" one).

dikrek wrote:

daniel_r328 wrote:

Wow bravo! The Fairchild adds a lot of legitimacy and shows how close the tweaks in the model come to the instrument in the video.

To me the reverb still gives it away - it's too clean. I'd use a shorter tail (or an external slightly more advanced reverb plugin?), and maybe experiment with adding random jitter to some of the reverb parameters (eg tone?). What I'd also do is introduce a noise floor to the sound (ie white noise track) to avoid pickup perfection. One thing I don't have experience in but am curious about is to add some artificial analogue saturation to the chain as well. It adds a kind of smear to ensure overlapping resonances aren't as cleanly separated.

I know you're not on Pro but for the record, I've found that raising the Hammer Tone to around 0.4 on the midranges makes them sound more convincing to me.

Thanks! Just added 2 more files, with 50% less hammer tone and 50% more (I can only affect the whole range but I'm curious how @dv will find that too).

Also removed the PTQ reverb and added a Lexicon 480L with the random hall mode, the tail should be - random

Plus some saturation with tape and more.

Which trends better now?

Used the sombre preset but modified. This time I'm not trying to match the recording really, but just to get a more accurate tone.

Could you try this MIDI file, Clair de Lune by Claude Debussy? It’s better balanced than your piece and could help you understand how to improve your sound in Pianoteq 9. https://forum.modartt.com/uploads.php?f...ebussy.mid

Added the result

This is also for daniel_r328, replacing that reverb with the QRS, which is supposed to blend the sound better with the source.

This I can confirm - QRS sounds more organic to me

daniel_r328 wrote:

The Fairchild adds a lot of legitimacy and shows how close the tweaks in the model come to the instrument in the video.

...

To my ears at least the "more hammer tone" option sounds like a step in the right direction!

Funny how we hear things differently. Of the two original ones, I like the one without Fairchild marginally better, but the difference is small.

To my ears, overall, the two new ones are much worse than the two previous ones. So my order of preference (and by that I mean "possibility to be mistaken for *ANY* acoustic piano", not "taste" or "similarity to the orginal") is

- dv_5th_NY_Steinway_Jazz_modified.flac
- Fairchild
- many other default presets
- MoreHammer
- LessHammer

In these new ones with more or less hammer, all the growling bass that made me so excited is completely gone. The piano tone is completely replaced with a muffled, distant version of itself. On the soft/quiet/pp, lyrical parts, the MoreHammerTone version marginally improves the attack, for example the section between 11' to 30' is marginally better. Neither of the two versions improve from 4'09" to 4'30" (in fact it's overall worse) but the problem there was not only the attack. On the relatively good soft part at 6'20-6'27" the LessHammer make it degrade and the MoreHammer changes it marginally but not for the better nor for the worse.

So, in conclusion, if you can do what you have done for MoreHammer to affect *only and exclusively* the "soft" parts, then yes, this will be a step forward. Otherwise the previous version was much better.

For what is worth, I'm judging all of these from a middle-of-the-road Philips Open headphones

Thanks!

This is like going to the optometrist!

OK, for you dv I did 2 more - they end in QRS_MoreHammerTone and QRS_LessHammerTone.

This is also for daniel_r328, replacing that reverb with the QRS, which is supposed to blend the sound better with the source.

Let me know folks!

I also recommend to try on a few different headphones and speakers, otherwise there's a high chance we're trying to optimize for a specific set of headphones or room/speaker combo...

The Fairchild adds a lot of legitimacy and shows how close the tweaks in the model come to the instrument in the video.

...

To my ears at least the "more hammer tone" option sounds like a step in the right direction!

Funny how we hear things differently. Of the two original ones, I like the one without Fairchild marginally better, but the difference is small.

To my ears, overall, the two new ones are much worse than the two previous ones. So my order of preference (and by that I mean "possibility to be mistaken for *ANY* acoustic piano", not "taste" or "similarity to the orginal") is

- dv_5th_NY_Steinway_Jazz_modified.flac
- Fairchild
- many other default presets
- MoreHammer
- LessHammer

In these new ones with more or less hammer, all the growling bass that made me so excited is completely gone. The piano tone is completely replaced with a muffled, distant version of itself. On the soft/quiet/pp, lyrical parts, the MoreHammerTone version marginally improves the attack, for example the section between 11' to 30' is marginally better. Neither of the two versions improve from 4'09" to 4'30" (in fact it's overall worse) but the problem there was not only the attack. On the relatively good soft part at 6'20-6'27" the LessHammer make it degrade and the MoreHammer changes it marginally but not for the better nor for the worse.

So, in conclusion, if you can do what you have done for MoreHammer to affect *only and exclusively* the "soft" parts, then yes, this will be a step forward. Otherwise the previous version was much better.

For what is worth, I'm judging all of these from a middle-of-the-road Philips Open headphones

Thanks!

daniel_r328 wrote:

dikrek wrote:

Added a second version with a Fairchild 670 compressor.

Wow bravo! The Fairchild adds a lot of legitimacy and shows how close the tweaks in the model come to the instrument in the video.

To me the reverb still gives it away - it's too clean. I'd use a shorter tail (or an external slightly more advanced reverb plugin?), and maybe experiment with adding random jitter to some of the reverb parameters (eg tone?). What I'd also do is introduce a noise floor to the sound (ie white noise track) to avoid pickup perfection. One thing I don't have experience in but am curious about is to add some artificial analogue saturation to the chain as well. It adds a kind of smear to ensure overlapping resonances aren't as cleanly separated.

I know you're not on Pro but for the record, I've found that raising the Hammer Tone to around 0.4 on the midranges makes them sound more convincing to me.

Thanks! Just added 2 more files, with 50% less hammer tone and 50% more (I can only affect the whole range but I'm curious how @dv will find that too).

Also removed the PTQ reverb and added a Lexicon 480L with the random hall mode, the tail should be - random

Plus some saturation with tape and more.

Which trends better now?

Used the sombre preset but modified. This time I'm not trying to match the recording really, but just to get a more accurate tone.

Could you try this MIDI file, Clair de Lune by Claude Debussy? It’s better balanced than your piece and could help you understand how to improve your sound in Pianoteq 9. https://forum.modartt.com/uploads.php?f...ebussy.mid

Which trends better now?

To my ears at least the "more hammer tone" option sounds like a step in the right direction! The reverb, too is a net improvement, but something I can't quite put my finger on is still off about it - it separates in my ear too much. Are you doing any sendback?

Sounds like you reined in the compression a bit and while this increases clarity, it pulls it a bit away from the video (but as you said, you're more interested in tone for now).

But DV's input will be more interesting as they have more perception around what's "off" with the raw tone

dikrek wrote:

Added a second version with a Fairchild 670 compressor.

Wow bravo! The Fairchild adds a lot of legitimacy and shows how close the tweaks in the model come to the instrument in the video.

To me the reverb still gives it away - it's too clean. I'd use a shorter tail (or an external slightly more advanced reverb plugin?), and maybe experiment with adding random jitter to some of the reverb parameters (eg tone?). What I'd also do is introduce a noise floor to the sound (ie white noise track) to avoid pickup perfection. One thing I don't have experience in but am curious about is to add some artificial analogue saturation to the chain as well. It adds a kind of smear to ensure overlapping resonances aren't as cleanly separated.

I know you're not on Pro but for the record, I've found that raising the Hammer Tone to around 0.4 on the midranges makes them sound more convincing to me.

Thanks! Just added 2 more files, with 50% less hammer tone and 50% more (I can only affect the whole range but I'm curious how @dv will find that too).

Also removed the PTQ reverb and added a Lexicon 480L with the random hall mode, the tail should be - random

Plus some saturation with tape and more.

Which trends better now?

Used the sombre preset but modified. This time I'm not trying to match the recording really, but just to get a more accurate tone.

Added a second version with a Fairchild 670 compressor.

Wow bravo! The Fairchild adds a lot of legitimacy and shows how close the tweaks in the model come to the instrument in the video.

To me the reverb still gives it away - it's too clean. I'd use a shorter tail (or an external slightly more advanced reverb plugin?), and maybe experiment with adding random jitter to some of the reverb parameters (eg tone?). What I'd also do is introduce a noise floor to the sound (ie white noise track) to avoid pickup perfection. One thing I don't have experience in but am curious about is to add some artificial analogue saturation to the chain as well. It adds a kind of smear to ensure overlapping resonances aren't as cleanly separated.

I know you're not on Pro but for the record, I've found that raising the Hammer Tone to around 0.4 on the midranges makes them sound more convincing to me.

dikrek wrote:

How different tastes are!

Does it mean you don't like that ? Or simply that it is very different compared to Debussy?

dikrek wrote:

Anyway, here's my attempt https://www.dropbox.com/scl/fo/almbpeap...2&dl=0

I don't quite know how to get the exact recording sound since these are mic and mastering choices I'd never make, but am I closer?

I did stuff like way increased hammer hardness and somewhat duplex, detuned stuff, more hammer and all other noises, and more.

Added a second version with a Fairchild 670 compressor.

I was not expecting the exact recording sound, and (perhaps because of my expectations being low) I find that you did a wonderful job in that regard, but I am not talking about those now, so let's just concentrate on the "acoustic-woody" vs "synthetic" sound.

I think you did the best job that I have heard with Pianoteq with the "growling bass"! By the way, is this v9? I haven't upgraded yet.... The incipit (first about 10 seconds) is simply fantastic and if I didn't know it was Pianoteq I could be fooled to think it was an acoustic piano, which is the first time that I experience that.

Unfortunately, the mid range lyrical parts have not that level of perfection. I mean, I like Pianoteq and they are acceptable to me, but I can see how they are different from an acoustic piano and how some people could hate them. Without having you to move to different times of the recording, the theme entering at the 11th second seem to be created by some sort of "wrong" hammers. Not exactly "too hard" but perhaps a bit "plasticy". The tone/timbre of the piano at that time is fine, it's just the attack that has that quality. I was actually tempted to add some hiss for fairness with the original recording to see if that made a difference, but I don't have any good source of hiss that I can tune.

These impressions are same for the whole recording: the busiest and more loud parts are really wonderful. The closest to an acoustic instrument that I've ever heard from Pianoteq (or any virtual piano, FWIW). In many places so close that's hard or even impossible for me to honestly say "I recognize it's a digital instrument and not an acoustic". All the soft parts though suffer from the problem above and in some places even the timbre itself is synthetic/plasticy. The worst part from that point of view is from 4'09" to 4'30" where I could clearly tell it's Pianoteq. Not just from the attack, but even the sustain is clearly "non-acoustic". In fairness you may have tried to closely match the original recording which I think suffer from the piano in that range being slightly out of tune (but still it's clearly recognizable as a slightly out-of-tune acoustic).

The best part of the lyrical ones you've got is the one from 6'20" to 6'27" which is still not perfect, but much, much closer to an acoustic than the examples I mentioned above (and the various similar parts of the piece). If you can make all the quiet/melodic parts like this one, it would be a massive improvement, and if you could even marginally improve those towards the "acoustic", you'd have reached the "holy grail" as far as I'm concerned.

But even if you can't improve that, I'm really impresse with what you've done so far! Hopefully you'll eventually share the FXP (or maybe sell it to Modartt )

Thanks a lot, really great job!

Thanks. It is v9 and I meant the recording isn’t to my taste at all (not the playing or the piece of course, who doesn’t like that! )

I’ll put a few other revisions and let you know when they’re done. It’s 0230 where I live now

How different tastes are!

Does it mean you don't like that ? Or simply that it is very different compared to Debussy?

I was not expecting the exact recording sound, and (perhaps because of my expectations being low) I find that you did a wonderful job in that regard, but I am not talking about those now, so let's just concentrate on the "acoustic-woody" vs "synthetic" sound.

I think you did the best job that I have heard with Pianoteq with the "growling bass"! By the way, is this v9? I haven't upgraded yet.... The incipit (first about 10 seconds) is simply fantastic and if I didn't know it was Pianoteq I could be fooled to think it was an acoustic piano, which is the first time that I experience that.

Unfortunately, the mid range lyrical parts have not that level of perfection. I mean, I like Pianoteq and they are acceptable to me, but I can see how they are different from an acoustic piano and how some people could hate them. Without having you to move to different times of the recording, the theme entering at the 11th second seem to be created by some sort of "wrong" hammers. Not exactly "too hard" but perhaps a bit "plasticy". The tone/timbre of the piano at that time is fine, it's just the attack that has that quality. I was actually tempted to add some hiss for fairness with the original recording to see if that made a difference, but I don't have any good source of hiss that I can tune.

These impressions are same for the whole recording: the busiest and more loud parts are really wonderful. The closest to an acoustic instrument that I've ever heard from Pianoteq (or any virtual piano, FWIW). In many places so close that's hard or even impossible for me to honestly say "I recognize it's a digital instrument and not an acoustic". All the soft parts though suffer from the problem above and in some places even the timbre itself is synthetic/plasticy. The worst part from that point of view is from 4'09" to 4'30" where I could clearly tell it's Pianoteq. Not just from the attack, but even the sustain is clearly "non-acoustic". In fairness you may have tried to closely match the original recording which I think suffer from the piano in that range being slightly out of tune (but still it's clearly recognizable as a slightly out-of-tune acoustic).

The best part of the lyrical ones you've got is the one from 6'20" to 6'27" which is still not perfect, but much, much closer to an acoustic than the examples I mentioned above (and the various similar parts of the piece). If you can make all the quiet/melodic parts like this one, it would be a massive improvement, and if you could even marginally improve those towards the "acoustic", you'd have reached the "holy grail" as far as I'm concerned.

But even if you can't improve that, I'm really impresse with what you've done so far! Hopefully you'll eventually share the FXP (or maybe sell it to Modartt )

Thanks a lot, really great job!