Closed captainmpc closed 9 years ago
Hum this occured not only at the beginning, and images append but the pixel buffer is incomplete :
good frame ( black area is a roto mask )
some bad frames
this one is really bad ( blocks and lines missing )
Ok, after some test, roto seems to be the bad guy. Disable it at merge node -> outputs ok Crop it before merging -> outputs more stable ( 4/5 OK) I try to replace roto with some random process ( blur/add/color correction ) -> outputs OK
So, roto rasterisation is a little buggy ;) Question : when roto is rasterized and which node provides the final raster resolution to it ? Request : menu action for cancel all current renders
This bug is similar to https://github.com/MrKepzie/Natron/issues/785 I'm closing.
Re-opening as this is in fact a different issue
I fail to reproduce this bug, can you send me the project? I do not necessarily need the input images, just replace it with a radial or constant
I do this now
can I send to you a wetransfert ? ( I package 100 frames with the natron project file ... maybe 5000x1420 image format cause this bug). email ? natron-contact ?
You can email me at alexandre.gauthier-foichat@inria.fr
No need for all that frames, just 5 is enough
On 29 Jul 2015, at 23:23, Sebastien notifications@github.com wrote:
can I send to you a wetransfert ? ( I package 100 frames with the natron project file ... maybe 5000x1420 image format cause this bug). email ? natron-contact ?
— Reply to this email directly or view it on GitHub https://github.com/MrKepzie/Natron/issues/786#issuecomment-126099301.
Maybe this can help. Roto node seems having a bounding box optimisation under the hood , but it show it !! When you scrub the time line, node output looks good. But if you zoom in as render area is bigger than view and you scrub , missing block and other glitches start to come in the cache.
Putting a crop node after roto one improve that.... even disabled ...
With "Pyramid" file. 2 viewers in graph but just one connected ( anywhere ). The other is left alone. Clear all caches. Scrub timeline. No pb with roto mask.
Now connect second viewer ( anywhere ). Clear caches. Scrub. Problem with roto like with more than one writer in action.
I try this with second test file ( green rectangle with holes ). Opened like the screnshot state ( two independent branch ). Problem when scrubbing. Ok. Not connected on same stream but problem is always here. So I delete one viewer. Clear caches. Bug still here ??!!
Because I notify another bug, I decide to reload default layout....two viewers tab appear ... the last viewer deleted is still here in memory... I delete it again. Save, close, load, just one viewer -> scrub OK.
Hope this can help you to find this bug. Roto rasterization is losted with more than one output .
I found the issue, now I have to come up with a solution and will update
Please confirm that this fixes it correctly
yes , it works in this snapshot ! thanks But in this one Natron-snapshots-75c49eff01f780e1bc70a076af44e73235235cdc.dmg , roto is broken ( no output at all )
last osx snapshot version
I'm working on videomapping project and I try to integrate Natron in my workflow. Here is my simple cropping script in aim to load five projectors with SVGA+ MJPEG videos outputs. Problem : when you render selected writers at same time, each result output file randomly flickers at the begin ( black frame blinking). It seems threads dont wait the processing branch before append frame in file.
Render each writers one by one works.
It is my bad ?
I try to enable "turbo mode", but render selected writers action ( dropdown menu ) do nothing at all in this case.