I think it may be harder to “sell” to the developers unless it was proposed as a package of measures e.g. Pro Fade Out, Cross-Fade, Fade Envelope (the “advanced fade”) and remove “Cross Fade In” and “Cross Fade Out”. Notice that there is no direct replacement there for a no-GUI Cross Fade In, which I still think is “unbalanced”.
Maybe you can persuade developers to have Pro Fade Out as well as Cross Fade Out and Cross Fade In for the time being, I don’t know. But if a major package of plug-in changes was on offer I don’t think the developers could ignore it as easily as a one off.
From my observations on the forum, the greatest demand (by far) is for some sort of “more versatile fade” effect.
Specifically, one that has:
Partial fade - fade to/from a level other than silence.
Some fade shape options.
These are also the features that I find most useful.
As a “general purpose” fade effect, I think that it needs to be reasonably versatile, but still reasonably simple. Something close to one of the “Fade by dB” versions would I believe make a good compromise.
Although I am able to tweak my installed fade effect to provide almost any fade shape imaginable, 99% of the time I use preset fade shapes.
I think that a limited selection of just these 5 fade shapes are likely to be adequate for the vast majority of users even without any “shape modifier” control.
As the “Amplify” and “Normalize” effect work in dB (rather than linear scale) I think that it would be sufficient for partial fades to be defined in dB rather than linear.
I don’t think that it would be necessary to have separate fade-in/fade-out versions as one plug-in could provide a drop-down option for fade-in or fade out.
The default could be “Fade In”, thus providing a balance against the “Pro Fade Out” effect.
The one other option that I think would be useful would be for the Fade-In/Out drop down to have options:
Options 3 and 4 provide a quick way to set two of the most commonly used fades and also allow “logarithmic” fades to or from silence.
(A true log fade will never actually reach silence, but hardware log taper faders get round this limitation by adding a short linear taper to silence at the bottom end of the slide.)
The “package” that I’m proposing is a “more versatile fade effect” (as described above) plus the “Pro Fade Out” effect, to replace “Cross Fade In” and “Cross Fade Out”.
Are there any feature that anyone would like to add or take away from the above specification?
Finally and most importantly - what will the “More Versatile Fade Effect” be called?
I would be very happy to support such a move, but I note that what we are lacking is a formal Proposal in the Wiki to develop such functionality, particularly the Fade Envelope/Advanced Fade (and I am not qualified or experienced enough to write such a proposal - though I’ll gladly work on the editing)
I was hoping too that you, Gale, would lead taking such a proposal to the developers, with support from Steve and myself, as you carry more weight with the developers that we two do. May we count on your support for this?
I am minded to agree with you there Gale - it would then be easier to argue for swapping out the existing Cross-fade-in and Cross-fade-out and replacing them with Pro-fade-in and Pro-fade-out.
But Steve’s tip of using multiple linear fade-ins is an excellent workaround producing good results - and I would be happy to upgrade the documentation/tutorials to recommend the use Pro-fade-out and multi-use fade-in.
Certainly the existing Cross-fade-in and Cross-fade-out effects are prime candidates for removal anyway as they don’t really “do what it says on the tin”. Steve’s newisg Nyquist effect cross-fade-classic does a much better job, but suffers from shortcomings imposed by the use of Nyquist and would be better properly implemented in C++ as part of the core code IMO and I would like to see this included in any proposal that gets written.
As I said above I was hoping that you, Gale, would lead this suggestion with our support.
And I totally agree that it would be better as part of a proposed package of changes (for which we would need a proposal as mentioned above). But this would then not preclude the devs from doing this in a step-wise fashion: adding Pro Fade Out first (giving a quick win) and then working on the other items in the proposal.
Yes, if I think it’s supportable, obviously. I would be prepared to put it as a Bugzilla enhancement if there was a Wiki Proposal to link to. That might help get it on to the agenda for a release. Steve is the resident fades expert, I think.
Equally sometimes one can just ask at the “right time” and it happens to fire a developer’s enthusiasm. However we seem to want to become much more formal of late with plans/agendas/things being “approved”.
Cross-fading built into Audacity could be a lot better than any Nyquist solution, and imho is ultimately the way to go, but that could take years to develop.
I would strongly prefer to not make implementing “Pro Fade Out” and “More Versatile Fade” dependent on implementing cross-fading as that is likely to hold the process for a long long time.
An alternative to offering 5 fade shapes would be to offer a smaller number (say 3) preset shapes and a “modifier” control (Mid-fade decrease or boost). For example:
Fade Shape (choice):
Mid-fade level adjust (+/- 100) <------------->
I think that these three presets are the minimum that need to be available as they are so commonly used.
Using the “Mid-fade level adjust”
Whichever preset is used, if the user wants the mid fade level to be higher, they would increase the “Mid-fade level adjust”. If they want to lower the mid-fade level they would decrease the “Mid-fade level adjust”. Examples:
An approximation to a logarithmic fade can be made by using the Linear fade and setting the “Mid-fade level adjust” to a negative amount (reduce the mid-fade level).
A fade shape that is more rounded than an Equal power curve could be created by using the Equal Power curve and increasing the “Mid-fade level adjust” to a positive amount (increase the mid-fade level).
Personally I don’t mind if we have the “5 options” or “3 options + modifier”.
What do you want? (question open to all).
That would not fully balance Pro Fade Out because there would still be no non-linear fade without GUI, even if the versatile fade default was Fade In.
I an broadly OK with the proposed content of versatile fade (see below). If there is room, I think a choice to define the fade by a percentage instead of dB is well worth having. About one-third of Feature Request votes for a better fade ask for a % choice.
I would still like to consider when we have a mockup of versatile fade if it should also support multiple envelope points.
I agree that it may not be essential to offer a Nyquist Cross Fade but I think we should perhaps have an idea what it would be if we were asked for it as the price of removing Cross Fade Out and Cross Fade In. I thought the Cross Fade you came up with that worked on adjacent audio was pretty good. As I recall, you don’t like the method that I feel is best for the least experienced users and I didn’t like your new mock up “simple” version which I felt left too much to be visualised in advance.
I would vote for
with the first three being mandatory, and “Round” not so important. I don’t mind a slider as well.
I am not clear to which of these options the dB (or %) control will apply. I think you mean that 3 and 4 will not take a dB control so as to provide a quick fade from silence to 100% and vice-versa, so if you want “from silence to 80%” you have to use Fade In? Again if there was a no GUI non-linear Fade In, this wouldn’t really matter, From Silence could still end up at other than 100%.
Also what are the default fade amounts or %?
Do you want a “Fade Out and In” and “Fade In and Out” choice as mentioned on the audacity4blind list?
Pass for now. That depends on its content and maybe that of the accompanying “package”.
Of course we still have the “fade to target dB” customers to satisfy, but I think I have resigned myself for now that will be a separate plug-in.
The Pro Fade Out effect has been optimised to provide a “musical” fade out effect.
For a “musical” fade in, in some cases a “positive” curve (starts steeply then levels out) will work best but in other cases a “negative” curve (starts gradually then more rapidly) will work better, so imho the optimum compromise is a linear fade in and we already have a no-gui linear fade in.
I think there should be no problem adding that. I’d prefer the default to be dB (like Amplify and Normalize).
Personally I think that will overcomplicate the plug-in as a mass market tool and (at this stage) my preference would be to make a multiple envelope point fade as a separate effect. According to feedback from Audacity4blind, that is the least useful feature of the Text Envelope.
If there is demand we could of course update the effect at any time and add that feature, but for now I’d prefer to just get something out there that works well. It is a lot easier to add a feature later than to remove one. Removing a feature (even one that does not work well) can be seen as regression.
I don’t see why it should be. “Cross Fade Out” and “Cross Fade In” are not crossfade effects.
I would much prefer to address a cross-fade effect as a separate issue. When we have a “more versatile fade effect”, that may well influence what features we want for a cross-fade effect, so let’s get this more easily attainable project done first.
BTW. Did we decide on “Cross Fade”, “Cross-Fade” or “Crossfade”?
The “Round” curve is the one that I use least often and if I were to choose one to leave out, it would be that one.
I have one more combination to offer to the table:
And then a “DJ Curve” slider that adjusts the curve of the “DJ Curve” option.
By default the DJ Curve would produce a fade shape that is approximately “Equal Power”, but as on the cross-fader (crossfader) of DJ Decks it has a “curve adjustment”.
(I like this combination).
Side note: The default “DJ Curve” could be identical to the current Cross Fade In / Out which would strengthen the case for it replacing Cross Fade In/Out.
Going from the feedback and from experience in using these effects for a good while, fading from or to 100% is what is used ~99.9% of the time. For the remaining 0.1% of cases users can apply the Amplify effect before or after the fade effect.
Again, if user feedback indicates that I am wrong about this, then we can always add a feature to allow setting both the fade from and fade to levels, but it would be harder to remove the feature (regression) even if no-one uses it.
Yes, though I don’t think that this is an essential option. The more I think about this, the less keen I am.
My preference would be that setting the slider to -100 dB goes to absolute silence. (this is a workaround for the lack of log scale sliders in Nyquist).
Not in this effect. I think that job is much better left to a specialised “multiple envelope point fade”.
I am not sure what you mean by the equal power curve. Normally, there are the typical curve types and a option which decides if the curves are applied as equal power or equal gain, no matter what the curve type is. For example: if we use a linear cross fade, the gain levels will be 0.5 for each signal in the middle of the transition. you could now generate two noise tracks of 5 seconds and apply this linear cross fade of 2 seconds. In the middle of the transition the signal will drop in loudness, regardless of the curve type you choose. This holds true for all uncorrelated signals and is called equal gain. thus the option for equal power. In practice, it means that we take the square root of the control signal i. e. the curve type prior to multiply it with the original signal. In the case of our linear signal this would mean:
First track 1 0.5 0 ==> 1 0.707 0
Second Track 0 0.5 1 ==> 0 0.707 1
As you can see, the two amplitudes in the middle of the transition add up to 1.414 (+ 3 db). This will counterbalance the perceived notch in the transition. However, if the two tracks are phase related (100 % correlation) the mix will clip in the middle since the max amplitudes (of one) will add up to 1.414. Thus, if you want to loop the same sound over and over again, the equal gain option is adequate. In real life, two signals are seldom perfectly correlated or uncorrelated. This is where the DJ-slider comes in. on the left would be the equal gain (0.5) and on the right side the equal power (0.707) i. e. a 3 db boost.
I think that the name equal power is therefore misleading as a proper curve name. Is it a cosine curve a gaussian or what? Or is it simply a linear line that is boosted in the middle?
Besides, Since we are talking about a non real time effect, could the volume correction not be applied half-automatically? only an idea.
Yes a linear fade, when used as a cross-fade is the equal gain fade
Yes the fade that is 0.707 mid-way is the equal power fade. It can not be linear.
There are a number of different curves that can produce equal power cross-fades. The one that I intend to use is a 1/4 cycle sine. As you say, this creates a 3 dB boost to the peak level mid-way through a cross-fade of uncorrelated noise.
Let’s see if I get it right:
For the equal power curve you multiply the fade-out portion with a 1/4 cosine and the fade-in portion with a 1/4 sine curve. at 1/2 Pi, the amplitudes will add up to 1.414, the square root of 2, don’t they. In contrast, the raised cosine goes over 1/2 of a period and the peak in the middle will be 0.5 in each signal, adding up to 1.
In a plug in that I wrote recently, the same types are available. But to make things easier, there is a additional option field, where you can choose between equal gain equal power and automatic volume control. In this way, it is possible to combine each curve type with any of the later 3 settings.
In order that the 1/4 cosine worked also, I simply squared it (0.707 * 0.707 = 0.5) for the equal gain mode. In power mode the result would be the same as your curve above. If you adapted a similar way, the DJ curve would fall away, whereas its slider would remain and gradually would go from equal gain (curve ^(1/1)) and equal power (curve ^(1/2)).
In other words, your plug-in would consist of a multi choice control with the different curve types and a real control with the extreme values for equal gain and equal power. This seems to me the most natural way. You may object that a linear curve type with a slider set to the far right would now be a curve with a bulge in the middle and no more linear at all. This is true, but the transition (for uncorrelated signals) would now be linear in the sense of loudness perception.
Clearly we could not have one “Fade Out” above the divider and another “Fade Out” below.
So is “Fade Out” (above the divider) “unmusical”, and what do you use for fading out speech?
“Pro-Fade” for the versatile effect is kind of OK, but it puts it a long way down the menu. I know it can have a shortcut, but you have to find the effect first.
Again I think this is a little futile (for the versatile effect) until we are sure what is in it, but I think its name should start with “F” or previous in the alphabet. I don’t see it as “pro” or “advanced” as proposed, it seems to be fairly “simple” but it’s a customisable fade that does fade in or out.
If it does not offer explicit envelope points as a control perhaps it should be called “Fade Dynamics…” or Fade Volume…", or even “Customised Fade…” (insert your word before “Fade”).
Yes, the existing “Fade Out” above the divider is un-musical compared with Steve’s new Pro-fade.
I don’t do much speech work, mainly music, so of course in the past I’ve always used the linear Fade Out as that was all that was available. However, I’ve just experimented on a bit of BBC-R4 and Steve’s Pro-fade seems to work absolutely fine on speech too to my ears.
I really would not want Pro-fade to remain below the line, it is too valuable a tool to sink down there; IMO it well deserves to be above the line, it would be more discoverable there.
For my money it could just replace the existing linear “Fade Out” - but I suspect that you may regard that as a regression (but we can work around that if the “Custom Fade” has a “linear” fade option as suggested).
I’m guessing that if we just replaced “Fade Out” with Pro-fade, calling it “Fade Out”, most folk would probably not notice. The difference is there but it is subtle and has to be listened for. But I do know that I, whbjr and PGA at least all prefer the Pro-fade.
All plug-ins are below the line.
The only ways that it could go above the line are through reorganising the Effect menu (proposal: http://wiki.audacityteam.org/wiki/Proposal_Menu_Reorganisation#Effect_Menu) or by coding it as a built-in Audacity effect.
It would probably not be hard to program as a built-in effect, but probably beyond my ability (I’m not a C+ programmer).
-1. (Linear) Fade In and Fade Out are fundamental effects for any audio editor.
For making a very short (milliseconds) fade out, the linear fade is better.
For cross-fades and drop-ins, the linear fade will probably be better.
For looping audio, the linear fade will probably be better.
One of the nice things about linear fade out is that you know exactly what you are getting - it’s simple.
If that is the case I, for one, would be very strongly “anti” replacing Fade Out. I use tiny fades (both in and out) to smooth off the end of cuts. It means I don’t have to worry about cutting at zero and I can get a totally click free join by
placing the two pieces on separate tracks and overlapping them by roughly the amount of the fades. I can then leave them like that or mix and render back to a single track.
So -1 from me to WC’s suggestion.