Page 3 of 6 FirstFirst 12345 ... LastLast
Results 31 to 45 of 84

Thread: Render Quality

  1. #31

    Not a "Little Bug"

    I must say that the implication that this is not an important issue seems odd.
    The main purpose of Speed Edit is to edit HDV footage and output a good, clean edit... a task which it is incapable of doing. This is a serious problem. This is a $500 product that doesn't do what it is supposed to do.

    I also wonder about Newtek... Newtek people have been able to comment on dozens of issues on this discussion group over the last two weeks.. often within hours... but not this one. A respect for ones cutomers would, it seems to me, require a comment... something like, "Wow, thanks for the testing we are checking into it." How hard would that be? The comment that they are too busy because of NAB? Well, this is one of their major items being shown at NAB... who will buy a product that doesn't work?

    I have been using Newtek products since Lightwave 1.0 and Amiga Toaster. They have been an innovative company over the years... and Speed Edit is an innovative product that I would like to benefit from... unfortunately I cannot now benefit from it and Newtek has chosen to ignore me and the other users represented in this thread. You be the judge.

  2. #32
    'the write stuff' SBowie's Avatar
    Join Date
    Feb 2003
    Location
    The stars at night are big and bright
    Posts
    19,376
    Quote Originally Posted by mlantern
    I must say that the implication that this is not an important issue seems odd.
    I don't think anyone is saying that it is unimportant. Just that it has only very recently been detected, and that at a time when all of the engineering team will be swamped with last minute work on stuff we here know nothing about.

    If you were to check back over the years you would see that - at peak pressure periods like this - (the normally near omnipresent) NewTek staffers are scarce as hen's teeth in the forums. There is nothing unusual or foreboding in this, and there can be no doubt the issue has been noted and will be addressed earnestly and expeditiously as soon as it is humanly possible. I suppose it will make everyone feel better when one of them does pop in to repeat what I just wrote, but it won't materially change the facts.
    --
    Regards, Steve
    Forum Moderator
    ("You've got to ask yourself one question ... 'Do I feel lucky?' Well, do ya, spammer?")

  3. #33
    Registered User KiloWatkins's Avatar
    Join Date
    Sep 2006
    Location
    New MexKilo
    Posts
    1,029
    Ahh.. I think I may understand the issue, more than the banter. HD @ 50Mbs may very well NOT show any of this artifacting. One of the main reasons we use Paint Programs for Print work. The HDV1080i .m2t's I have printed to tape on the source HDV Sony, just jump off the screen with color and clarity, viewing back on an IMac and Hi-End Samsung 32".

    YMMV
    GoodLuck KiloVideo

    Digitizing video since 1984
    HyMPG Designer since 1997
    Avatar says, 89.3 MPG Avg.
    [email protected]

  4. #34
    Media Tech FSU robertn2k's Avatar
    Join Date
    Feb 2003
    Location
    Tallahassee, Florida
    Posts
    93

    Render Quality

    Well...I'm sorry if you received the impression from my email that this is a “trivial” matter. It certainly is not, but I still find raking Newtek over the coals and suggesting that we start threatening to return the product in mass a bit much, especially when tech support has had the issue for a little over three business days. I've reported things to Adobe and it's going on 90 days and I still have not had any acknowledgement from them. The fact that Newtek does respond to this community pretty quickly should not be a reason to condemn them just because they have not responded on this one issue.

    I have a chain of command in my business and if there is a tech problem it gets verified by my staff and then a tech person before it reaches me. I would assume Newtek has something similar. I think once Dr. Cross get's his hand on this we will have a response and a solution pretty quick. Has anyone emailed Dr. Cross directly about this issue with their test results?

    Respectfully,

    Robert


    [QUOTE=mlantern]I must say that the implication that this is not an important issue seems odd.

  5. #35

    Dr. Cross

    I'm sorry, but I do not know who Dr. Cross is... nor do I have his email. If you think he can help and you have his email I would appreciate it if you would email him and ask him to read this thread... if he hasn't already.

    So, I thought to myself, "I'll email tech support about this and maybe they will get in touch with Dr. Cross." Unfortunately, when I went to "email tech support" Speed Edit was not listed in the drop down box of products. Did I miss it? I looked again and still couldn't find it. I guess I need some help here also.

    For everyone who thinks that the upcoming NAB show is a good excuse I wish to point out that NAB doesn't even open until 16 April... that's still 21 days away. If nobody at Newtek does anything until after NAB it's going to be a long wait.

    Finally, I'm just a customer trying to figure out how to make their product work... and I would appreciate some help. Several users, who at first doubted the problem, did their own tests and have verified the problem... look at the posted test files... and I thank them. And I thank anybody inside Newtek or outside who can contribute positively to this discussion.

  6. #36
    Paintball Video Geek billmi's Avatar
    Join Date
    Jul 2004
    Location
    Palm Bay, FL
    Posts
    1,773
    Quote Originally Posted by mlantern
    I took the original .m2t into After Effects and rendered an uncompressed AVI at 1920x1080. The results were the same--render quality better than Speed Edit and equal to an 8 bit 422 Black Magic uncompressed render.
    If I understand this right, you are comparing After Effects output rendered to an uncompressed format to Speed Edit rendered to a compressed format (Speed HQ.)

    If you are going to compare apples to apples, shouldn't both renders be to the same output format, in order to eliminate the possibility that the loss is inherrent to the compression of the output format, rather than the engine dealing with the image before it gets compressed?
    Toasting with NewTek since 1991 - Corinthian Media Services
    Riding Quiet on an Electric Ninja.

  7. #37
    Paintball Video Geek billmi's Avatar
    Join Date
    Jul 2004
    Location
    Palm Bay, FL
    Posts
    1,773
    In a few minutes of playing with Cholo's test pattern - and outputting to a .tiff still with varying quality levels in its control tree - it looks like the test pattern is being resized slightly to fit the project.

    In low quality some of the black lines disappear. In medium and high quality they get blurred like they are being anti-aliased.
    Toasting with NewTek since 1991 - Corinthian Media Services
    Riding Quiet on an Electric Ninja.

  8. #38
    Paintball Video Geek billmi's Avatar
    Join Date
    Jul 2004
    Location
    Palm Bay, FL
    Posts
    1,773
    Cholo - your test pattern is 1920 pixels wide. Speed Edit's default configuration for 1080i is 1440 pixels wide. I changed it to 1920, loaded in your test pattern and saved a still with all the vertical lines intact and looking correct. There was still an issue with the horizontal lines, and I haven't tried any tweaks to the vertical size, but this really looks like an issue of scaling/rescaling to me.
    Toasting with NewTek since 1991 - Corinthian Media Services
    Riding Quiet on an Electric Ninja.

  9. #39
    Visually Developed Monkey cholo's Avatar
    Join Date
    Feb 2003
    Posts
    636
    billmi

    I was careful not to resize anything during my tests. The tests I did with hdv material were done on a 1440x1080 interlaced timeline with 1080i hdv material. All my test renders went out to uncompressed and were transformed into jpegs outside speededit (in virtualdub). Comparison shots were hdv streams imported directly into virtualdub, saved as uncompressed avi and then reimported and saved as jpeg so each workflow had the exact same number of steps involved. The tests I posted using the test pattern were done on a full res 1920x1080 progressive timeline and again rendered out as uncompressed avi, imported into virtualdub and converted into jpeg for posting here. I was very careful with my testing procedures.

  10. #40
    Paintball Video Geek billmi's Avatar
    Join Date
    Jul 2004
    Location
    Palm Bay, FL
    Posts
    1,773
    Cholo,
    I should correct my last post to say that 1440 width is the default for HDV 1080i. In playing with scales some more, with Quality set to low, the issue with Cholo's pattern really looks related to scaling to me - lines getting dropped during a rescale, and getting smoothed to a grey if medium or hiqh quality mode is used. And no, I'm not trying to say you're throwing things at it in wacky resolutions, but that perhaps Speed Edit might be doing something scaling wise when it shouldn't be, or shouldn't need to. The artifacting looks like every nth line thrown away to shrink an image by 1/n when set to low quality, and the pattern of missing lines changes accordingly with slight scale or position shifts.

    mlantern - I noticed on the sample images you posted, they were all in high or medium quality. I know this sounds counter-intuitive, but did you try low quality? The less processing that gets done to a clip, the closer it should be to the original. High quality and ISS do great jobs of keeping things pretty when rotating and scaling, but shouldn't be necessary for unmanipulated video. That and of course outputting to an uncompressed format.
    Last edited by billmi; 03-26-2007 at 04:27 PM.
    Toasting with NewTek since 1991 - Corinthian Media Services
    Riding Quiet on an Electric Ninja.

  11. #41

    Chart Test

    Dear Billmi:
    Thanks for joining the discussion. I have not previously posted any tests... several people have and they are similar to the "real" image tests I initially did.
    Cholo's are particuarly interesting because they represent both real world images and charts.... and because Cholo initially didn't think there was a problem, but through careful testing was convinced.

    I did want to post a chart test I did because you can clearly see the difference in resolution as well as the blurring of horizontal lines but not vertical lines. The chart was shot with a Sony FX1. The file was imported into both Premiere and Speed Edit via Firewire... However both files looked the same. The file was put into a 1440 HDV project in both pieces of software and rendered to a 1440 uncompressed AVI. The avi's were exported to bmps in Premiere and converted to jpgs in Photoshop. Kind of round about, but Premiere didn't want to do jpgs directly. The resulting jpgs are, however, a good representation of the actual render quality and I might point out the renders are a good representation of how things look on the monitor.

    Now I have done this test in Speed Edit at all possible combinations of layer settings all the way from low to high with and without SS. I have done the renders with all possible variations of field setting and I have done the renders to .m2t, Quicktime, Black Magic, Uncompressed AVI and probably a couple of more things along the way. I have done the renders in Speed Edit, Premiere Pro native HDV, After Effects, Premiere Pro Cineform, and Premiere Pro Black Magic. I have done 1440 and 1920 renders as well. I have tried to slightly scale the image in Speed Edit to see if that made a difference. Interestingly if you change a setting... say you scale to 101% the image from Speed Edit comes into focus and then goes out of focus in about a 1/2 second. In all cases Speed Edit is the only program that blurs the image. I hope this provides some clues to anyone interested in this discussion or problem.

    I don't seem to be able to attach those jpgs... I'll try again later. However, they clearly show a horizontal resolution wedge that goes out to 600 lines with the lines well differentialed all the way out in the original and the Premiere render, but merging at around 450 lines in the Speed Edit render.

    It is interesting that in the past two weeks many have posted similar results, but no one has posted renders that are "good." Hope to hear from you again and anyone else interested in this product.
    Bob

  12. #42
    Visually Developed Monkey cholo's Avatar
    Join Date
    Feb 2003
    Posts
    636
    I've posted good renders at SD resolution. In fact, SpeedEDIT has the best scaling engine in any NLE when downscaling HD material to SD. It's only when working at full HD (at and beyond 1080) that problems do surface. In fact, 720p appears to be fine in terms of quality when downscaling 1080 24p material. Here's a nice image to see how it matches up quality wise to the VirtualDub lanczos rescale filter, considered by many to be the absolute best quality method for downrezzing HD material
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	AvengerVdub0000.jpeg 
Views:	197 
Size:	466.8 KB 
ID:	44549   Click image for larger version. 

Name:	AvengerSE0000.jpeg 
Views:	193 
Size:	437.2 KB 
ID:	44550  

  13. #43
    Visually Developed Monkey cholo's Avatar
    Join Date
    Feb 2003
    Posts
    636
    And a color corrected version out of SpeedEDIT just for fun!
    Attached Thumbnails Attached Thumbnails Click image for larger version. 

Name:	AvengerSEcc0000.jpeg 
Views:	209 
Size:	479.1 KB 
ID:	44551  

  14. #44
    XDCAM HD production ScorpioProd's Avatar
    Join Date
    Feb 2003
    Location
    Albany, NY
    Posts
    4,346
    OK, question...

    I can certainly see that this is an issue when one is dealing with something like CGI or video for film output or any high-end HD formats at 1080i resolution...

    But with the heavy compression already in HDV, is this something that would be actually noticable with HDV2 1080i?

    More importantly, should I not be editing any of my 1080i HDV in SpeedEDIT till this is addressed if I want maximum quality?

    Cause honestly, if I'm shooting in HDV, my intent is higher quality, so I do find this thread rather worrying. Especially since HDV support is really one of the main reasons to use SpeedEDIT for any of us that already have VT-EDIT.
    Eugene
    Scorpio Productions

  15. #45

    1080i Versus 720p some good news

    Dear Cholo:
    I think you have discovered an important detail. When I look at what is happening in 1080i the problem sure looks like something associated with how Speed Edit is reading and rendering the fields. When I looked at Uncompressed AVIs from Premiere and Speed Edit at 1080i, 1080p, and 720p my results agree with yours. 720p from Premiere and Speed Edit are almost identical (this is from an original 1440 m2t). In fact, the 720p render is significantly sharper than the 1080i render from Speed Edit. This lead me to consider 1080p in Speed Edit. My initial test indicates 1080p is actually the same as 720p in Speed Edit... they are hard to tell apart.

    The conclusion seems to be that Speed Edit is having problems with the fielding associated with 1080i or 1440i... or maybe more accurately, Speed Edit is having trouble outputing the fielding associated with 1080i. So this makes 720p an acceptable HD output mode for Speed Edit... significantly better than 1080i while 1080p seems to offer no improvement. This is good news... not the best news, but good news. Thank you for your effort.

    By the way, how do you post those jpgs? May "attach" button doesn't seem to work.
    Bob

Page 3 of 6 FirstFirst 12345 ... LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •