RSS feed reader

98 Guy wrote in :

And ditto for "Pale Moon", which is quite similar to FF, but was
written specifically for Windows, and is not cross-platform,
like Firefox (it's kind of an offshoot of Firefox, IIRC).

I've messed a little with Palemoon, but found that I got the white-line
across bitmapped images while scrolling a page up and down (just like in
FF 3.x) so I didn't pursue it.


If there is any 'smooth scrolling' or similar control, does the problem
persist if you switch it off?

Also, does it improve if you use a keyboard instead of a scrollwheel to
scroll? (ignore that if you don't use a wheelmouse...)

98 Guy wrote:
Bill in Co wrote:

And ditto for "Pale Moon", which is quite similar to FF, but was
written specifically for Windows, and is not cross-platform,
like Firefox (it's kind of an offshoot of Firefox, IIRC).

I've messed a little with Palemoon, but found that I got the white-line
across bitmapped images while scrolling a page up and down (just like in
FF 3.x) so I didn't pursue it.

I've also tried a browser called "Avant" - which if I recall is based on
the IE rendering engine. At the time I had an older version of
KernelEx, and it almost worked (it seemed to fully load, but it then
crashed the system). If anyone is so inclined, and has the time, might
want to try it and see if they can get it working.

Yes, as I recall Avant was based on the IE engine, whereas Orca was not, at
least for the versions I tried. (all this on the XP computer). I'm not
sure you'd gain all that much by getting Avant to run, even if you could,
however.

I also tried installing Pale Moon 3.5, but it seemed to mess up my FF 3.5
setup as I recall, so I scrapped that (on the 98 computer).

I finally went back and tried FF 2.0 (from the backup image) and installing
an updated Flash (in my case, just to ver 10.3), but ran into more
difficulties than I did with FF 3.5 in doing this, of all things.

I think part of the extra difficulty was that FF 2.0 was even older than FF
3.5, so that added yet another layer of BS to wade through, since Flash ver
9 is the latest you can use on Win98 without KernelEx and some workarounds
(one of those I had to use was to disable the plugin compatibility checks,
in about:config, etc).

Plus I didn't like seeing that annoying "your browser is out of date"
message on some web sites. So I guess I'll leave it with FF 3.5.8.

Lostgallifreyan wrote:
98 Guy wrote in :

And ditto for "Pale Moon", which is quite similar to FF, but was
written specifically for Windows, and is not cross-platform,
like Firefox (it's kind of an offshoot of Firefox, IIRC).

I've messed a little with Palemoon, but found that I got the white-line
across bitmapped images while scrolling a page up and down (just like in
FF 3.x) so I didn't pursue it.


If there is any 'smooth scrolling' or similar control, does the problem
persist if you switch it off?

Also, does it improve if you use a keyboard instead of a scrollwheel to
scroll? (ignore that if you don't use a wheelmouse...)

I don't recall seeing some of those scrolling problems he's mentioned before
with FF 3.5.8 (similar to Pale Moon), so it may be at least partially unique
to his setup. Then again, I haven't spent tons of time investigating it.

Lostgallifreyan wrote:
"Bill in Co" wrote in news:Wf-
:

I have even sometimes used Print Screen to get a better printout of
what's
on the screen for a few websites (i.e. paste image and print image from
within Wordpad).


I use the PrtScr button a lot (and Alt+PrtScr for small windows.) The
tools
that take long page screenshots basically do the same thing, copying a
bitmap
from the screen buffer, based on some co-ordinates. They just scroll the
page
enough to get a new sample, and splice the samples together. It's just
that
browsers use nonstandard scroll controls that aren't so easily controlled
as
native controls, and 'smooth scrolling' is especially bad because the
image
must settle before the copy is taken. Sample size has a big impact on the
performance of the tool, as does memory handling. I might code one because
it's a very good test of several important principles, and I never found
one
that worked as I wanted it. And there may never be one that works for
everything. At those times, manual PrtScr and splicing our own large
images
is the only way. I built up a large local map once, by first getting a
load
of Google Map satellite image tiles for high res, batch-shrinking and
sharpening them for better visibility, then scripting a large HTML file
that
tiled the output files to rebuild the map in a browser. I then used
PicPIck
to scroll it and make long vertical strips. I finished by splicing each of
those in PaintShop till I had a huge single image. I used them in a
progream
I made in wxLua, called ChartGPS, which shows a track in a way that lets a
single glance determine speed and direction, overlaying a very detailed
image
of the terrotory. It's ideal for cross-country run mapping, but would
easily
scale up to hiking, biking, sailing routes, hot air balloons etc. I will
probably try to port it to C and native API at some point.

Re click removal, I always do it by hand, as the results are second to
none,
but I only do it when the audio justifies it. I can't see so well now so I
might not do it again unless I use a screen that has larger pixels. It
took
ten times as long as some auto-tool, but it was like comparing 10% with
99.999999%, there really is no contest. Small means nothing when you have
the
fast view and zoom controls of Sound Forge. One moment it's like viewing
the
ground from a jetpack, the next you're on it like a scanning tunelling
microscope. Scale changes really are no obstacle, it's just a question of
whether it's worth the time. I have imagined processes that might help
automate it, but it's partly integration in the time domain by eye, and
partly differentiation in frequency domain by ear. Combing those in code
well
enough to beat the human brain might be enough to win a Nobel prize for
advances in forensic examination and augmentation of human senses by
machine
aid, and even if in the unlikely event I manageed it, it would take longer
to
run on my machines than doing it myself!

Well, in the case of at least some of the audio files I've collected over
the eons of time, there is no way I could do some of them by hand (not with
thousands of small clicks), unless I wanted to spend a year doing it. :-)
The results from using "Click Repair" (in the automatic mode) can often be
pretty good in those cases - as long as your only dealing with small clicks.

I agree with you that the larger clicks must be done manually, as we've
discussed before. And fortunately, it usually works out that there are a
LOT fewer large clicks in most material, so doing it manually is no biggie.
:-)

"Bill in Co" wrote in
m:

Well, in the case of at least some of the audio files I've collected
over the eons of time, there is no way I could do some of them by hand
(not with thousands of small clicks), unless I wanted to spend a year
doing it. :-) The results from using "Click Repair" (in the automatic
mode) can often be pretty good in those cases - as long as your only
dealing with small clicks.

I agree with you that the larger clicks must be done manually, as we've
discussed before. And fortunately, it usually works out that there are
a LOT fewer large clicks in most material, so doing it manually is no
biggie.
:-)



I could risk going over old ground, but there's another tack I never really
explained.. There are things I hear sometimes on BBC's Radio 3, restored old
recordings from shellac disks, etc, classical music recordings, where I know
that I could have done a better job removing the surface noise and reducing
the noise floor without harming the detail and the 'feel' of the recording by
smoothing it out too much. The reason that bothers me is that they have
access to stuff like Cedar (probably mis-spelled there). So the question is
what are they missing? My guess is that the best automation really doesn't
come close to a good cleaning by brain, eye, ear and hand, followed by a bit
of noise reduction and harmonic regeneration once the noise has been cleaned
to the point where it is mostly peridic or at least not discontinous.

I guess the reason they don't do it is that they can't find anyone they trust
to do it, and pay for the time it takes. That;s what is so strange, with some
of those recording, I'd have thought money was no object, it's like restoring
'old master' paintings. In their case, lasers are used, but the same applies,
a lot of careful operation by eye, brain and hand. No substitute for it has
ever been found, but we do at least have a few more tools to help with it.

Weirdly enough, it's actually the smaller clicks that I find work best by
hand. Large ones are often easy to spot with tools, but they often make a
clumsy cut. The tricky bit is telling a large range of small spikes from
distortion, as opposed to the atonal component of a bowed string sound, for
example. On the really small scale, I can tell those spikes apart, and I have
cleaned sustained soprano voice sounds of bad styleus-induced distortion. The
best tools I know of can't do that, they either make the voice lose its edge,
(or do similar over-smoothing on violins), or they leave a high noise floor
that includes a large residual amount of distortion or vinyl surface noise. I
have easily, and often, got better results than on many recordings the BBC
proudly claims as restored. But I don't know why, other than the guess I made
in this post. I think it is just that no-one ever spent long enough at it
manually to convince those who were going to pay, that the result might be
worth it.

We live in an age of machines, where people try to avoid the work that the
old masters had to do. The truth about restoring is that we too must use
their methods, not the ones that come too easily to us.

"Bill in Co" wrote in
news
I don't recall seeing some of those scrolling problems he's mentioned
before with FF 3.5.8 (similar to Pale Moon), so it may be at least
partially unique to his setup. Then again, I haven't spent tons of time
investigating it.


Maybe. I just metioned smooth scrolling because it's an example of something
that can change a line in ways that conflict with attepts to render its
pixels in more detail. The fact that it IS a line suggests scrolling anyway,
and most blank line errors I have seen were implicating a scroll method in
some way. Snap scrolling, the old and easy and fast way, is best, because
anythign more complex demands time, CPU power, risks poor rendering, etc. A
lot of web pages seem to impose their own (maybe JavaScripted) methods to
influence scrolling, so the ideal might be to fins a master override.

Lostgallifreyan wrote:
"Bill in Co" wrote in
m:

Well, in the case of at least some of the audio files I've collected
over the eons of time, there is no way I could do some of them by hand
(not with thousands of small clicks), unless I wanted to spend a year
doing it. :-) The results from using "Click Repair" (in the automatic
mode) can often be pretty good in those cases - as long as your only
dealing with small clicks.

Because large clicks must be "removed" or cleaned up manually, I should add.

I agree with you that the larger clicks must be done manually, as we've
discussed before. And fortunately, it usually works out that there are
a LOT fewer large clicks in most material, so doing it manually is no
biggie.
:-)



I could risk going over old ground, but there's another tack I never
really
explained.. There are things I hear sometimes on BBC's Radio 3, restored
old
recordings from shellac disks, etc, classical music recordings, where I
know
that I could have done a better job removing the surface noise and
reducing
the noise floor without harming the detail and the 'feel' of the recording
by
smoothing it out too much. The reason that bothers me is that they have
access to stuff like Cedar (probably mis-spelled there). So the question
is
what are they missing? My guess is that the best automation really doesn't
come close to a good cleaning by brain, eye, ear and hand, followed by a
bit
of noise reduction and harmonic regeneration once the noise has been
cleaned to the point where it is mostly peridic or at least not
discontinous.

And one thing that must always be kept in mind here in audio restoration is
the maxim, "Less Is More". (which is true for so many things :-)

Many "restored" recordings that I've heard have been over-processed, and
it's really annoying. Those folks that did it obviously don't get it.
:-) You have to leave in some of the imperfections to do this well.

I guess the reason they don't do it is that they can't find anyone they
trust
to do it, and pay for the time it takes.

Right. Or figures there's not enough demand to make it worth it, at least
for most of the public.

That's what is so strange, with some
of those recording, I'd have thought money was no object, it's like
restoring
'old master' paintings. In their case, lasers are used, but the same
applies,
a lot of careful operation by eye, brain and hand. No substitute for it
has
ever been found, but we do at least have a few more tools to help with it.

Well, I'd say restoring a painting might "register" with more folks, so
that's why they do it. That's just "more visible" to most folks.

Weirdly enough, it's actually the smaller clicks that I find work best by
hand. Large ones are often easy to spot with tools, but they often make a
clumsy cut.

Yes, trying to repair a large click can be difficult. And if you listen
carefully, you'll often hear some vestiges of it, which makes perfect sense,
since more material was compromised. But I usually resort to just using
FSE (freq space editing) tools to clean it up (which fill it in the F-D, by
looking at the adjacent regions), which can (often) do a pretty reasonable
job, rather than by manually drawing it in (as I expect you normally do).
If I've got a lot to work on, I'd much rather do it this way.

Interestingly enough, on noise reduction, I often will just use a flat noise
profile, but in really bad cases, I will take a noise sample and use that.
Why would I use a flat noise profile? Because I get less artifacts that
way, and a flat and unaltered frequency response contour, obviously, unlike
with using a noise profile contour, which, of course, alters the overall
frequency response of the processed material.

The tricky bit is telling a large range of small spikes from
distortion, as opposed to the atonal component of a bowed string sound,
for
example. On the really small scale, I can tell those spikes apart, and I
have
cleaned sustained soprano voice sounds of bad styleus-induced distortion.
The
best tools I know of can't do that, they either make the voice lose its
edge,
(or do similar over-smoothing on violins), or they leave a high noise
floor
that includes a large residual amount of distortion or vinyl surface
noise. I
have easily, and often, got better results than on many recordings the BBC
proudly claims as restored. But I don't know why, other than the guess I
made
in this post. I think it is just that no-one ever spent long enough at it
manually to convince those who were going to pay, that the result might be
worth it.

We live in an age of machines, where people try to avoid the work that the
old masters had to do. The truth about restoring is that we too must use
their methods, not the ones that come too easily to us.

I don't mind doing some work, but I'm not going to spend a month removing
thousands of small clicks manually. :-) (Well, unless someone wants to
pay me to do it by the hour. :-)

The programs I can't stand are the "all in one" audio restoration programs
that (supposedly) can do ALL the cleanup work (noise, clicks, crackles, and
what have you) for you by pressing a button. THOSE are terrible. :-)

"Bill in Co" wrote in
m:

Yes, trying to repair a large click can be difficult. And if you listen
carefully, you'll often hear some vestiges of it, which makes perfect
sense, since more material was compromised. But I usually resort to
just using FSE (freq space editing) tools to clean it up (which fill it
in the F-D, by looking at the adjacent regions), which can (often) do a
pretty reasonable job, rather than by manually drawing it in (as I
expect you normally do). If I've got a lot to work on, I'd much rather
do it this way.


We're pretty good at interpolating, so i use that fact. Given a large gap, I
won't try to fill in the full frequancy domain in any way at all, unless I am
certain that a very contant repetition is held there. This works with solo
brass and other similar timbres, but not with a lot else, especially strings.
It's ok with many synthesized or processed sounds, maybe.. The reason is that
the HF signature is complex. Even if I had FSE I'd have to try to merge the
content to avoid subtle and complex discontinuities. What I do instead is to
look at the average shift in the time domain, it's like looking at a
landscape, you can see where the hill contours are despite bushes and trees
cluttering the horizon, and where someone left a gravel pit or stuck up a
spiky tower, it's easy to gauge the lie of the land. I won't always just
leave a smooth curve either, especially if the gap persists for many tens of
samples, I'll wiggle the line by hand, while following the contour, to
simulate the more erratic undulations that modulate that curve. If it looks
wrong, I undo and retry,and I test by ear too. This sounds crude, but it
works, it uses mind in a way no machine can work. If it feels right, it
usually sounds right too. Trying to copy the whole HF spectreum into that gap
will often fail, because it conflicts with expectations. A crude but accurate
reduced-info fill in is better because the mind will do the rest! It really
is effective. For similar reasons, a mind will hear flutter and wow in a tape
machine, but it is far less sensitive to dropouts, mostly because they rarely
involve HF transients.

Interestingly enough, on noise reduction, I often will just use a flat
noise profile, but in really bad cases, I will take a noise sample and
use that. Why would I use a flat noise profile? Because I get less
artifacts that way, and a flat and unaltered frequency response contour,
obviously, unlike with using a noise profile contour, which, of course,
alters the overall frequency response of the processed material.


I agree that a flat profile works, and that may be like the above stuff, it's
based on reducing real problems, but does not add details that conflict with
a brain's expectations. I usually go a tad further (as I do when adding
undulations to a curve in a gap left by large-scale transient removal). In
this case I use a noiseprint that is close to the content I'm fixing. I have
a generic FM stereo noiseprint that rarely needs retaking (though I advocate
making a new one any time you change an antenna or receiver type or
location). I have a vinyl one too, but again, I'd change that if I used a
different turntable or changed its mat.

About the all-in-one tools, agreed, they're fairly horrible. I'll be polite
about them because even if I let rip my real feelings it wouldn't deal with
them. About the time taken to do it manually, I had a vinyl copy of Toyah's
song Ieya, a 12" recording, before there were any CD releases I knew of. All
the long held voice sounds were ravaged by stylus distortion, over about 5
minutes of a song lasting 8 minutes and 20 seconds. I tried all sorts of
stuff, then gave up and had at it by hand because at least it workd, and the
time spent thrashing around looking for any kind of tool was taking a couple
of days. Once I'd done it by hand it had taken me two sessions over two days,
total about 12 hours, less time than I'd spent finding tools that wouldn't
come anywhere close to working. It's quicker than you might think, just
tedious, tense, maybe like playing that game where you have to pass a small
loop along a winding maze of bare wire without it touching, and with the wire
long enough to make it a scary prospect, but I found it like tightrope
walking (yes, tried it..), you just maintain a basic awareness of context,
remain stable, and keep going. I liked the process, but I agree, I'd rather
get paid if I have to do it often.

Lostgallifreyan wrote:
"Bill in Co" wrote in
m:

Yes, trying to repair a large click can be difficult. And if you listen
carefully, you'll often hear some vestiges of it, which makes perfect
sense, since more material was compromised. But I usually resort to
just using FSE (freq space editing) tools to clean it up (which fill it
in the F-D, by looking at the adjacent regions), which can (often) do a
pretty reasonable job, rather than by manually drawing it in (as I
expect you normally do). If I've got a lot to work on, I'd much rather
do it this way.


We're pretty good at interpolating, so i use that fact. Given a large gap,
I
won't try to fill in the full frequancy domain in any way at all, unless I
am
certain that a very contant repetition is held there. This works with solo
brass and other similar timbres, but not with a lot else, especially
strings.
It's ok with many synthesized or processed sounds, maybe.. The reason is
that
the HF signature is complex. Even if I had FSE I'd have to try to merge
the
content to avoid subtle and complex discontinuities.

I'm not sure if you could do that other than letting the FSE do the blend.
You can customize some options on that blend, but I don't think you can go
in and do what your saying in the FSE editing window.

What I do instead is to
look at the average shift in the time domain, it's like looking at a
landscape, you can see where the hill contours are despite bushes and
trees
cluttering the horizon, and where someone left a gravel pit or stuck up a
spiky tower, it's easy to gauge the lie of the land. I won't always just
leave a smooth curve either, especially if the gap persists for many tens
of
samples, I'll wiggle the line by hand, while following the contour, to
simulate the more erratic undulations that modulate that curve. If it
looks
wrong, I undo and retry,and I test by ear too. This sounds crude, but it
works, it uses mind in a way no machine can work. If it feels right, it
usually sounds right too. Trying to copy the whole HF spectreum into that
gap
will often fail, because it conflicts with expectations. A crude but
accurate
reduced-info fill in is better because the mind will do the rest! It
really
is effective.

I think you'd find that if you actually tried out some FSE editing, it might
surprise you. Actually, I think it would work better in many cases than
hand editing a large gap! As you said, you might have to wiggle lines by
lots of repeated trial and errors in the TD to get it to "sound better".
But that step is in essence automatically done for you and more accurately
(in the FD, however) by the FSE blending (which takes in account the
surrounding material). And I think, much more accurately than one could
ever hope to do by randomly using wiggle lines in the TD to crossover the
gap, which is a totally blind approach.

For similar reasons, a mind will hear flutter and wow in a tape
machine, but it is far less sensitive to dropouts, mostly because they
rarely
involve HF transients.

Interestingly enough, on noise reduction, I often will just use a flat
noise profile, but in really bad cases, I will take a noise sample and
use that. Why would I use a flat noise profile? Because I get less
artifacts that way, and a flat and unaltered frequency response contour,
obviously, unlike with using a noise profile contour, which, of course,
alters the overall frequency response of the processed material.


I agree that a flat profile works, and that may be like the above stuff,
it's
based on reducing real problems, but does not add details that conflict
with
a brain's expectations. I usually go a tad further (as I do when adding
undulations to a curve in a gap left by large-scale transient removal). In
this case I use a noiseprint that is close to the content I'm fixing. I
have
a generic FM stereo noiseprint that rarely needs retaking (though I
advocate
making a new one any time you change an antenna or receiver type or
location). I have a vinyl one too, but again, I'd change that if I used a
different turntable or changed its mat.

About the all-in-one tools, agreed, they're fairly horrible. I'll be
polite
about them because even if I let rip my real feelings it wouldn't deal
with
them. About the time taken to do it manually, I had a vinyl copy of
Toyah's
song Ieya, a 12" recording, before there were any CD releases I knew of.
All
the long held voice sounds were ravaged by stylus distortion, over about 5
minutes of a song lasting 8 minutes and 20 seconds. I tried all sorts of
stuff, then gave up and had at it by hand because at least it workd, and
the
time spent thrashing around looking for any kind of tool was taking a
couple
of days. Once I'd done it by hand it had taken me two sessions over two
days,
total about 12 hours, less time than I'd spent finding tools that wouldn't
come anywhere close to working. It's quicker than you might think, just
tedious, tense, maybe like playing that game where you have to pass a
small
loop along a winding maze of bare wire without it touching, and with the
wire
long enough to make it a scary prospect, but I found it like tightrope
walking (yes, tried it..), you just maintain a basic awareness of context,
remain stable, and keep going. I liked the process, but I agree, I'd
rather
get paid if I have to do it often.

98 Guy wrote:
Bill in Co wrote:

And ditto for "Pale Moon", which is quite similar to FF, but was
written specifically for Windows, and is not cross-platform,
like Firefox (it's kind of an offshoot of Firefox, IIRC).

I've messed a little with Palemoon, but found that I got the white-line
across bitmapped images while scrolling a page up and down (just like in
FF 3.x) so I didn't pursue it.

Maybe you should try FF 3.5.8 (not 3.6) once again just for a test ride and
see if you still get the same thing, as I don't recall seeing it. PLUS the
other poster said he'd heard of the same thing happen sometimes with FF 2.0,
at least as I recall.

The advantage of going to FF 3.5 is that it should work better on more sites
nowadays, although even it is pretty dated. So FF 2.0 is getting a bit long
on the tooth, just like IE6 is. (I still have IE6 on the Win98 computer but
it's limited in which sites I can still access properly anymore). And it's
all just going to get worse, I'm afraid.

I bet the day will soon come that unless you have a HTML5 compatible
browser, you'll be unable to access properly a lot of sites. But I'm really
not looking forward to that, as I'm getting tired of this "pretty much
forced on us" upgrade crapola. MS already did that with IE6 (can't run
anything higher on Win98 assuming you want to use a MS browser, which you
probably don't).

I've stopped at XP, and expect it to stay that way, for years to come. But
then again, I also spent countless hours (months?) beating and customizing
XP into submission, LOL. So it "looks" pretty much like my Win98SE
computer now, but with the added bonus of being able to run a few nice apps
that I can't even get to install and/or run on Win98SEeven with KernelEx
(and even at that, they are fairly old editions - I generally don't like the
newest bloatware stuff)