Any other poor souls using PFClean?

Perry Paolantonio Feb 8, 2016

  1. We bought this last year because it did one type of fix incredibly well, but it's been an unmitigated disaster otherwise. I feel like it's the 1990s and I'm using Mac OS 9 software: random crashes, corrupted project files, weeks of lost work, etc. The developers have basically told me (when we crash the software) "Don't do that." I've also pointed out to them that it's impossible to complete a single reel of restoration on a feature film without maxing out the system's memory (we had to upgrade from 32GB to 64 to work around this), and was told it's a low priority.

    Dealing with their support folks has been an exercise in futility. It takes 24-36 hours to get a response, if you get one at all. there doesn't seem to be any sense of urgency, but there's plenty of time to fiddle with the user interface in ways that are utterly meaningless, when the underlying software is such a house of cards.

    We can now reliably crash the system, hard, in 4-5 different ways. Do they care? It really doesn't seem so.

    In any case, if you want to commiserate, I'm here and I feel your pain.
     
  2. That completely sucks. Have you looked into replacements like MTI, Algosoft, or Diamant?
     
  3. We switched from MTI actually, which I've used for over a decade in various incarnations. We'll probably wind up going back to that as a primary tool, because it's a great tool for lots of things, and one of the biggest workflow advantages is that it's doing work directly on the files you import, as you apply the fixes, rather than requiring a render, or generating terabytes of temporary cache files. It's super efficient. With PFClean, a 1TB 4k scan (one reel of a 35mm feature) requires at least 3TB of disk space: 1TB for your source, 1TB+ for your cache files, and 1TB for your final render).

    But MTI is not very good at automated cleanup work, and dealing with the specific issue we were dealing with (fingerprints on film that were only in the blue channel, for four or five frames on either side of every single splice) just became too time consuming. PFClean was able to address that problem in seconds.

    I've looked at Diamant before and found it to be kind of funky, but perhaps it has improved. We're also considering getting Algosoft, which I used when it was first released (or possibly in beta - it was a while ago), and the interface left a lot to be desired: no manual tools at all, and a UI only a Windows 95 programmer would love. But it was a pretty cool automated tool, and is on our list as a potential standalone dustbusting tool.

    I realize the answer here is to have multiple tools and to play to their strengths, but with restoration work, I don't want to have to manage multiple versions of projects (automated pass done in one tool, and then manual fixes in another, etc), because backing out of changes or making changes becomes much more difficult that way. With PFClean, the fixes are done in an effects stack, so you can easily back out of one item in the stack at any time, or make adjustments or whatever. However, it's so poorly implemented that if you just look at the software the wrong way, it'll crash...
     
    Marc Wielage likes this.
  4. My first job out of school was working on PFClean and Pre-BMD DaVinci Revival. Also the first job that ever skipped out on the bill, but that's another story.

    We never had any trouble with it, but I was only an operator so I don't know anything about the setup. This was in Late 2011
     
  5. Jason Myres

    Jason Myres Moderator

    3,628
    2,267
    0
    We have a PFClean at work as well, and we have one operator who's taken to it and really likes it. I have not heard him complain of crashes, but I know he has had to have it rebooted a few times. It seems, like grading, each restoration system has a few things it's excellent at, and a few others it totally misses. Apparently none of them cover it all.

    Since you're already very familiar MTI, have you ever given Phoenix a try?
     

  6. We're interested in Phoenix as well. I've sat through some demos at NAB and I'm impressed, but it was always priced way out of our range. I've been too busy to call back the sales guy to get current pricing though, and they don't seem to want to email me a quote. I'll have to get on that soon.

    Like I said, we have found that PFClean has some really strong points. I do feel like they're spending a lot of time on user interface stuff though, and not addressing underlying problems. Someone on the PFClean google group has suggested that we try it on a mac. I'd be more open to that if I was sure it'd be more stable, but tech support has told me I won't really see a difference. Is it better on a mac? I don't know. I'd prefer to keep it on Windows or even Linux, because we can keep the machine in the rack (so it doesn't take up floorspace), and so that we can upgrade it when necessary. I've got no interest in the new macpro's, because it's a terrible design that locks you into what you buy, or into buying overpriced Thunderbolt interfaces for otherwise cheap connections to basic stuff.

    Some of the issues, like using up all the memory in the system, are there by design according to their tech support folks. They claim they have to "strike a balance" between memory usage and performance. That's fine. If the system requires 128GB of RAM to work correctly, and I knew that up front, I'd have built that. But they say on their site that you need 8GB of RAM and a 1GB GPU. We now have 64GB and a 6GB 980ti in there, and we're still having tons of problems.

    When we scan a film, we do it by reel, and that's fine - we'd bring each reel into PFClean and restore them separately. but sometimes we get complete films as files, and with MTI we could load that whole thing in and just run with it. With PFClean, they're telling us we need to break up our 20 minute reels into smaller pieces, to avoid memory issues. I mean, that right there is just bad design.
     
  7. We used a Digital Vision 2K DVNR which is a realtime signal based cleaner thingy. It did a good job with automatically cleaning dust and scratches. It obviously can't do any bigger fixes like cleaning splices or splotches or broken frames so it was a semiautomatic workflow. The Phoenix uses the same tools and if I was really for realzies making a bunch of restoration work I'd probably get that. If I had the money for it, that is.
     
  8. Our workflow with MTI was always to do everything manually, no automatic cleanup at all. That made sense with MTI, because its auto-pass tools aren't very good. But it's fantastic with manual cleanup, making some really seamless fixes with almost no effort.

    PFClean has some really great auto-dirt tools, and we're making extensive use of them on the current project, which is a real mess. But those tools seem to be responsible for the excessive memory use so we're stuck. I like this hybrid workflow, where we run an auto-pass first and then hit the whole film manually with manual fixes. It's a lot quicker than what we've always done, and with the quality of fixes we get from PFClean, it's a good solution -- when it works, which isn't often.

    So if Phoenix can do something similar, I'm open to that. I'd much rather have everything in one system than having to bounce stuff around, because I like being able to revisit a project in 5-6 years and know exactly what we did. Invariably, we need to do that. If it's been through half a dozen pieces of software, it's much harder to track what we did to what.
     
  9. I bet some of the memory issues would go away if you just broke it down to 20-minute reels, then put them all back together at the end. You're never going to be able to process more than 20 minutes of material in a day anyway, assuming there's lots of pos & neg dirt plus splices plus warps plus vertical bumps plus rips. To me, this is the best answer.
     
  10. Hi Marc,

    So we are working in 20 minute chunks (most of the time), and even that's too much for PFClean in some cases. We sometimes work with complete features and MTI handled that just fine, but with anything we scan ourselves, it's always done by reel. What the Pixel Farm people are telling us is to break the 20 minute reels into even smaller chunks. The problem with this is that a single film goes from having 5 separate projects to having 10-20-30 separate projects, to deal with all the little pieces. It becomes a logistical nightmare, and it clogs up the UI. It's workable if we're only doing one film at a time, but we usually have 3-5 restoration jobs going on at once, all with different deadlines and schedules. Some are long-term, lower priority jobs that we work on as we have free time, others are more urgent.

    -perry
     
    Marc Wielage likes this.
  11. Oh, well, then that's totally stupid. Have you called these guys in England and yelled at them? This is preposterously wrong. I knew the DRS guys at Technicolor a few years ago (most of whom are now laid off, sadly) and they typically worked in 20-minute chunks and did OK. I know a company using Pixel Farm on the East coast; I'll contact them and see what they're using and if it works.
     
  12. The general attitude from The Pixel Farm seems to be that they have our money and they're not really interested in making this work. Look, it may very well be a configuration issue on our end, but all evidence (and that includes anecdotal evidence from other users) is that the software is built on a very shaky foundation and this kind of problem is the norm.

    I think it's telling that the bug report form (which is built into the application) includes a section where you can detail a workaround, if there is one. When you look at some of the open bugs on their support site, many of which have been there for a long time, they do point out workarounds. But to me that's the wrong approach - they're acknowledging the bug exists, but just telling you how to avoid it, instead of fixing the underlying issue. Several of these bugs are crashes, and are listed as lower priority than I would assign a crash. Also, there are only maybe 50 bugs shown (which I find hard to believe - I've worked on much less complex software projects that had hundreds of open bugs at any given time), and most haven't been addressed in the 6 months we've owned the software. This, despite weekly builds being available for end users to download.

    I worked in the software world for a long time, so I know how to track down and report a bug. Anything that causes a crash should be an automatic Medium or High priority bug at minimum (I'd say High to Critical), but they seem to think if there's a way to avoid it, it can be put off. I think there's a systemic problem with the way The Pixel Farm does their development, and that extends out to the tech support. Also, I'm fairly certain they're a small company so odds are the programmers are the tech support and QA people, and that's not usually a good thing.

    Seems one solution might be to revert back to the 2013 version (assuming our license allows that). That version came out before they started messing with the UI and people seem happier with it. Of course, none of the projects we've done in the past 6 months would be openable if that's the case, because all that work was all done in PFClean 2015.
     
    Marc Wielage likes this.
  13. Reading about all these trials & tribulations I really think you should try Phoenix. Their support staff is solid and have a helpful attitude, a far cry from what you are describing.
    There are some complex restoration problems the software can't solve because they are more in the area of advanced compositing but for 98% of your tasks it does them very well and conveniently. Auto-dirt, auto-scratch, auto-flicker, auto-steady are reliable and clean (few false positives) and the manual dirt tool allows you to work quickly. Good UI and very stable.
    It does need to render all auto-effects for real-time playback and depending on how many tools you stack you can have 16 cores on a Xeon workstation at 100% for the night. Then in the morning you review, make corrections and do the manual work on top of that. A structured workflow pays off.
     
  14. So at this point, I've spent the holiday weekend working on getting a project exported. I was able to figure out a series of workarounds on my own with zero help from tech support until today. And the help I got boiled down to this:

    "I'm afraid at this point in time we cant offer any advise on getting to the root cause of the problem, all we can offer is trips/tricks [...] Until we get to and address the real root cause of your issues you'll probably continue to run into problems I'm afraid and all we will be able to offer in the short-term is band-aids and workarounds."

    In other words, "figure it out yourself and then we'll look at it." No suggestions are made as to a starting point, just a blow-off, pretty much. At least, that's how I read this.

    How does one deal with this? It's not like we're talking a $250 piece of software here. For a small or even mid-sized shop, $10k is a substantial outlay (not to mention we've lost at least 5x that over the past few months in re-doing work due to file corruption, etc).

    Obviously we're looking for alternative software here, but others seem to be able to use PFClean successfully (though many are on mac and Linux). We're toying with installing linux on that machine and seeing if it's more stable there, but that's going to require a fair bit of time to set up and we can't do it mid-stream when we've got 3 different restoration projects going at once. So who knows when I'll be able to try that out.

    What a disaster this has been.
     
  15. Thanks for being open about this. I think we all need to hold vendors feet to the fire if they make a product that doesn't preform as it should. The margins are so tight today that "bad" purchases aren't easily absorbed.
     
  16. I'm now being told that they will revisit this *after* they get their 2016 release out the door, but that in order to do it, they need to reproduce the problems. That means I guess we'll have to send them the entire 2TB project/cache file plus 1TB of source media for one of the reels, in the hopes that they can see the instability I'm seeing. At least that's something, I guess.

    It's just not a priority because they have come up with this kind of ridiculous annual version cycle, rather than just releasing a version when it's fully ready. By most accounts, from users who have been on PFClean for a while, the 2014 and 2015 versions are kind of unfinished software, and even in the 6 months we've had 2015, it's been through one drastic UI change in the media management. Documentation is out of date with the current software, showing UI elements that no longer exist, etc.

    I spent the weekend in the office and managed to get this current project exported out, and now it's back in Resolve to finish it off. But it shouldn't have been nearly this complicated. Part of the promise of PFClean is that it allows you to do a lot of that stuff in one application (exporting to various resolutions/frame rates, aspect ratios, etc). But I just don't trust it enough to mess with that at this point.
     
  17. Any progress on that front?
    I have just posted in a thread about the config for PF Clean, wondering if things had got better.
    Software looks good but it's a lot of money if it is not able to deliver
     

Share This Page