Flowframes Early Access Download: 1.36.0
Added 2022-01-02 15:26:59 +0000 UTCThis update comes with the RIFE 4.0 model which improves quality over 3.9 while also being faster.
In addition, some bugs have been fixed.
Full changelog on Discord and in Flowframes.
Comments
Thanks!
Renaud Z.
2022-04-10 10:54:32 +0000 UTCYes, you can check my Discord for details, but a new version is ready soon. Within one week most likely.
LTSM
2022-04-10 10:24:30 +0000 UTCHello, Hi, This little message to find out if this project is still active and if it's still useful to support it financially, since it's been three months without news... Thank you for the news !
Renaud Z.
2022-04-09 21:07:17 +0000 UTCPretty sure this odd behavior is because I have a 12700k with its E Cores. -Python.exe "default normal priority" max cpu usage is 50%, while changing it to "above normal priority" max cpu usage is 70+% for that process (normally 50% because FFMPEG takes up 50%). -FFMPEG.exe "default below normal priority" max cpu usage is 20%, "normal priority" cpu usage is 50%", and "above normal priority" gives 70+% usage, but normally 50%. -While Flowframes window is foreground I can get 100% cpu between the 2 processes, but if I click on anything else (background usage) the max cpu usage is 20% for both processes. -Python, decoding, encoding, and clean up are massively faster with the higher cpu usage. FFMPEG.exe decoding at 20% cpu usage is 6 fps , while 15 fps at 50% usage.
Bryan Metros
2022-02-24 00:47:25 +0000 UTCYeah XVFI works much better in matter of quality, but I wasn't able to interpolate 4K on my 3070. I am always getting error that I need more VRAM. I can interpolate 720p.. RIFE works on 4K without any issues.
Vojtech Novak
2022-02-23 16:55:44 +0000 UTCAh damn, is there any possibility in the future? I'm having much better results with the XVFI model than RIFE (less artifacting, more detail, smoother etc).
Rajiv Lutchman
2022-02-21 13:17:26 +0000 UTCThis code currently doesn't support multi-GPU.
LTSM
2022-02-21 12:57:41 +0000 UTCI'm currently interpolating some 4K frames using the XVFI model and no matter what I set in the dedicated GPU options (0,1,2 - in both fields), only one of my 3x3090 GPU's is utilised for the interpolation. Can anyone help please?
Rajiv Lutchman
2022-02-21 12:48:10 +0000 UTCselect "60fps" at "Maximum Output Frame Rate" in "Quick Settings" page
hihi
2022-02-13 12:23:55 +0000 UTCHi there. Please add exact framerate select as an option. For example, My video is 25FPS, and I only want to convert to 60FPS. But there are only 2x, or 3x... So I have to choose 3x and my video will become 75FPS. It takes longer render time and I have to convert it to 60FPS again after completed. Thanks.
Long Cao
2022-02-05 03:21:08 +0000 UTCWhy does this version take so long to download? On the install page I've been waiting for it to load for hours but I can't wait any longer. I tried changing the server and it didn't help any faster. Help me, I remember in the previous version I downloaded much faster.
jobasun
2022-02-01 09:38:45 +0000 UTCThe software would be excellent if it wouldn't use the hard disk so much. It probably slows down the process and decrease the life time of my SSD. Is there no way to compute as a "sliding" task (using only the RAM) instead of a step-by-step (using huge amount of disk ressource) ?
Vincent
2022-01-31 10:58:46 +0000 UTCHey n00mkrad first and foremost thanks for the incredible tool you've created. I do have an issue related to speed. I have a laptop equipped with a 10750H and rtx 2060 and my encoding speed is about 2.5~FPS when encoding a 7 minute 4k clip. Is that normal? Is there anyway to speed that up maybe twice as much if not more with my current system?
Josh Richardson
2022-01-26 18:38:30 +0000 UTCThis worked, it downloaded from vmd78749.contaboserver.net now and actually made some progress. :) So I guess there are 2 pieces of feedback here: 1. The installer stated that the download size for the model preload was 400MB, but that may be outdated info in the installer since the model cache size says it's 2.3GB. That makes more sense to me 2. Selecting USA as the download location during the installer may actually be selecting auto rather than specifically selecting USA. Thanks!
Chris Mayfield
2022-01-15 10:26:50 +0000 UTCCloud providers have strict traffic limits and usually don't allow direct links. You can manually select the US server, see if that works for you. (Settings -> Developer Options)
LTSM
2022-01-15 09:17:56 +0000 UTCI selected the USA location for pre-downloading the models which it said were 400MB, but it connected to static.242.177.55.162.clients.your-server.de which does not resolve to a USA host (or point to the IANA reserved IP address in the name) and proceeds to download from 162.55.177.242 which is nmkd-hz.de. The network transfer was going fast enough that it should have blown past 400MB in no time, but it kept going seemingly forever. :/ Why not host the model files on one of the cloud providers for free or cheap? That said, you've done a decent job of streamlining these. It's a lot less hassle than having to set everything up on a new machine and dig into docs just to get things configured right. :)
Chris Mayfield
2022-01-15 08:48:42 +0000 UTCDude... I already sent 2 messages on Discord, but I got no answer. Could you please think before sending such useless messages ?
Renaud Z.
2022-01-14 12:34:51 +0000 UTCDM me your Discord name on Patreon if the invite doesn't work, possible you were banned for whatever reason, but the Patreon comments are not great for support
LTSM
2022-01-11 16:07:20 +0000 UTCStill no news from N00MKRAD...
Renaud Z.
2022-01-11 15:57:07 +0000 UTCThank you Alice.
Renaud Z.
2022-01-11 15:56:55 +0000 UTCI'm having the exact same issue - the operation is automatically cancelled, saying the extraction of frames failed and the input file is incompatible, even though the exact same files interpolated with 1.35
Alice
2022-01-11 15:27:32 +0000 UTCWhen I click on the Discord icon, i get an "invite invalid" error. Anyway, the full error is described below.
Renaud Z.
2022-01-04 21:55:13 +0000 UTCJoin the Discord by clicking the Discord icon in the main window, go to the channel #flowframes-support, upload it there with a description of the problem. Patreon messages are a bit too clunky to provide efficient support.
LTSM
2022-01-04 21:21:52 +0000 UTC[0] [1-4-2022 21:8:18]: Command Line: "C:\Users\xxx\AppData\Local\Flowframes\Flowframes.exe" [1] [1-4-2022 21:8:18]: Files: None [2] [1-4-2022 21:8:18]: Args: None [3] [1-4-2022 21:8:19]: Initialized Nvidia API. GPU: GeForce RTX 3080 [4] [1-4-2022 21:8:19]: [Servers] Ping to nmkd-hz.de: 33 ms [5] [1-4-2022 21:8:19]: SymlinksAllowed: True [6] [1-4-2022 21:8:19]: [Servers] Ping to nmkd-cb.de: 119 ms [7] [1-4-2022 21:8:19]: [Servers] Closest Server: Germany (Nürnberg) (33 ms) [8] [1-4-2022 21:8:19]: Parsing Patrons from CSV... [9] [1-4-2022 21:8:19]: Found 13 Gold Patrons, 363 Silver Patrons [10] [1-4-2022 21:8:20]: [UI] You are running Flowframes 1.36.0. The latest Patreon version is 1.36.0, the latest free version is 1.31.1. [11] [1-4-2022 21:8:20]: [GetGpus] Found GPU: Intel(R) UHD Graphics 630 [12] [1-4-2022 21:8:39]: Selected video/directory: Spectacle 1920x1080 (GHQ)p25.mov [13] [1-4-2022 21:8:39]: [UI] Your file seems to be on an HDD or USB device. It is recommended to interpolate videos on an SSD drive for best performance. [14] [1-4-2022 21:8:39]: [UI] Loading metadata... [17] [1-4-2022 21:8:40]: Getting frame count (F:\DVD\movie576p25.mov) [18] [1-4-2022 21:8:40]: Hash not cached, reading frame count. and then, Flowframes shows "loading metadata" but nothing happens...
Renaud Z.
2022-01-04 20:13:21 +0000 UTCAnyway, the (short) log is here :
Renaud Z.
2022-01-04 20:12:32 +0000 UTCHow can I send a log on Discord ?
Renaud Z.
2022-01-04 20:12:18 +0000 UTCThanks, I was hoping it wasn't buried in the appdata folder!
Kezo
2022-01-04 05:24:34 +0000 UTCCan't reproduce, please send logs on Discord
LTSM
2022-01-03 20:49:49 +0000 UTCYou can set the temp folder location in the settings, by default it's in the directory of the input video
LTSM
2022-01-03 20:31:01 +0000 UTCHi, There is apparently a regression since version 1.35: the framerate detection is no longer done ("0 FPS" remains), and the operation is automatically canceled after the extraction of the frames. Tested on several videos and several formats (Windows 11, RTX 3080). The same videos, on version 1.35.00, are processed without problem.
Renaud Z.
2022-01-03 20:14:48 +0000 UTCAlso, why is PyTorch on such a slow server? It's taking forever for a 1Gb file, and I'm on a gigabit connection! Never mind, I had to stop it, and did a restart, now the connection is much more tolerable, though still slow as hell!
Kezo
2022-01-03 02:11:10 +0000 UTCWhere is the cache folder located? As I have a QVO SSD and I want to limit the number of writes that occur on the drive...
Kezo
2022-01-03 01:11:23 +0000 UTCNICE! Overjoyed to take a look at the new model!
Kezo
2022-01-02 23:55:59 +0000 UTCHuh. I've talked with various other people about how we use SVG for VR interpolation since FF always messes up the aspect ratio. I'll try again on the new version.
baxtyr
2022-01-02 20:14:56 +0000 UTCI have interpolated VR videos in the past, works just like any other video, just slower :p
LTSM
2022-01-02 19:58:52 +0000 UTCis there any way to interpolate VR media yet? I havent been able to figure it out. Always gets squished
baxtyr
2022-01-02 19:43:43 +0000 UTCAwesome. Look forward to trying it out
spriggsyspriggs
2022-01-02 15:30:45 +0000 UTC
