[Kdenlive-devel] Title Widget / Monitor scene speed

[Dan Dennedy]

On Tue, Aug 24, 2010 at 11:09 AM, Till Theato <root at ttill.de> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 08/24/2010 07:53 PM, Simon A. Eugster wrote:
>> On 21.08.2010 11:29, Dan Dennedy wrote:
>>> On Sat, Aug 21, 2010 at 1:27 AM, Simon Eugster<simon.eu at gmail.com>  wrote:
>>>> Regarding the title widget, I (as already stated before) vote for a
>>>> rewrite anyways. Perhaps we can get rid of performance problems as
>>>> well then.
>> [...]
>>> BTW, in thinking about masks (better term than "region"), I am
>>> thinking about a way to modify MLT to allow one to place special
>>> filters "Set mask" and "Apply mask" onto the filter stack. That way
>>> one could designate one set of filters to apply with a mask, another
>>> set with another mask, and others with no mask. The mask defined in
>>> "Set mask" would be used with all filters below it and above "Apply
>>> mask." And "Apply mask" clears the mask for filters below it until
>>> another "Set mask." When the "Set mask" filter is selected, it could
>>> expose the "mask monitor overlay widget" mentioned above.
>>
>>
>> Regarding masks. I would love to have a widget that allows me to select a
>> region on the monitor to take the color information from for the scopes.
>>
>> Sometimes I just want to select a neutral area in the monitor, or maks out
>> part of the clip to check whether it was this part causing some points in
>> the blue area of the vectorscope.
>>
>> Don't know what would be suited best here.
>
> If you add an option to your scopes to ignore transparent pixels, this
> could already be done with for example the frei0r filter Alpha Shapes.
> And of course if I get to the point of writing a polygon masking filter
> (mentioned in this thread) this could be done more precise.

True, and it can useful to apply filters only temporarily for
analysis. However, it is important to understand that in MLT, the mask
will be an additional channel to the alpha channel. However, these are
always distinct parts of the MLT Frame and not necessarily its image
data that gets fed into QImage.

Previously, I mentioned that I have to provide a way to preview the
different branches of the filtering (pulling vs. applying). Instead of
compositing over black - as mentioned in my other reply - I could make
a preview option that takes the image prior to the start of masking
and simply copies the mask channel into the alpha channel and then
that turns into a QImage. That would be slightly faster than
compositing with black. Following Till's advice, it would give you a
short term solution (temporarily apply an alpha filter) as well as
something that will continue to work when this filter masking rolls
out.

-- 
+-DRD-+