Flowframes Early Access Download: 1.39.0

Please update your Flowframes app

Romuprod

Hello, do you have any update for me? Thanks

Gianni

During the installation process I select auto for everything. Please advice since I'm new to this. Thanks

Gianni

Please consider I have 2 GPUs the ids 0 is Intel Iris Xe 1GB ram, and ids 1 is GeForce RTX-3080 ti with 16GB. if I left 0 in settings - AI Framework settings both for Pytorch GPU ids.. and NCNN Gpu ids it is slow but it completed the process. If I change it to 1 for both to let software uses the RTX-3080 it is fast but suddenly it generate that error. I tried many times with same result. Please advice, thanks

Gianni

This is what I get https://we.tl/t-VkUv9WMpjs

Gianni

Yes, should work fine with a 3080 Ti

LTSM

Hello, do this version support GeForce RTX-3080 ti? Every time I set this as primary cards in the software settings I got error

Gianni

i am by the way on 1.36.0... maybe 1.39 fixes this or is this an inherent issue with interpolation?

Raf Stahelin

hey, loving the speed of the interpolation on rifre4. I have a a video that has a tiled pattern for background. there's a bit of wobbliness on the lines on the interpolated frames. how does one mitigate this jelly effect. It is subtle but still noticeable thanks. here's by the way a link to the original and interpolated mp4's https://www.dropbox.com/sh/27apqqvxhto048z/AAD58alDWJ_fjAX7RFrll1vya?dl=0 thanks

Raf Stahelin

Rife 4.6 came out almost 2 months ago. You gonna update anytime soon?

Upscale Anon

Any update on the next update? are you taking a break?

stephen

No, SDR (8-bit only)

LTSM

Dose this version supports converting 4K/HDR/24p to 4K/HDR/60p?

Ed

works great here, zero issues. Problem is probably on your end

Pewburrito

ŦɭคยЖєɭเภ

there is a bug with the batch processing queue, if you load multiples folders by drag and drop in the batch processing queue, it won't fetch the input Fps from what was entered in the interpolation tab. adding a "default fps" in the options would be wonderful :)

Charlie Leroy

same for me :(

Charlie Leroy

any update?

Riccardo Fasoli

Using a folder of PNG files as input. For some reason Flowframes "extracts" the frames from this PNG folder prior to interpolation. I.E. it just makes a copy of the same PNGs. This takes a long time. Any way to make it just reference the original PNGs for the frames? Thanks.

Rici

The queue is broken in this version. Dragging and dropping files in to the queue causes the first file to run through but open reaching the second file I get a "Microsoft.net framework" unhandled exception "DragDrop registration did not succeed".

Liam

Wow... What a marvelous piece of software. Thank you.

LustGard

Love this app, I've been using different AI upscaling and interpolation tools since working on the old Magic Lantern hack for Canon DSLR's. This is probably one of the best I've tried. Excited to see what's next.

KingSamuel1st

Thank you for the solution! The installer appears to function as usual now but thats a useful workaround!

Joel Morris

I disabled defenders firewall for the install and turned off SSL certificate checks (https://www.youtube.com/watch?v=Xp5G8x3SNhE), worked for me.

Pepelaugh

Yeah same happens to me

Pepelaugh

Hey, I seem to be getting a "the underlying connection was closed" error, is there a solution anyone knows of?

Joel Morris

That's rather unlikely. But you can just encode losslessly to avoid quality loss

LTSM

Hello N00MKRAD, thank you very much for Flowframes! I use it to restore digitized childhood Super8 film. The digitization gives me a 1440x1080 video stream at 18 fps, which I upscale to 2880x2160 60fps. Thanks for the fractional speed increase! Now I do not have to go 18 fps -> 180 fps -> 60 fps. Since the restoration does, however also involve (AI-) Upscaling/Compression artefact reduction and subsequent stabilization steps, this currently involves an unsatisfactory number of re-encodes with the inherently accumulating encoding losses. Do you see a way to enhance Flowframes to include also the AI upscaling/artefact reduction (such as in Cupscale) or even camera jitter stabilization in a single process without intermittent re-encoding? For me, this would be a game changer and I believe there is current no single solution that is really up to this task. From my experience, the task sequence upscaling -> frame interpolation -> stabilization yields the best and most consistent results. Thanks again for your great work!

Trantor

Use RIFE NCNN VS

LTSM

I am getting "Neither the Flowframes Python Runtime nor System Python installation could be found!" even if Python 3.10.7 is installed and in my PATH environnement variable... Any suggestion?

Sylvain Ménard

Hi N00MKRAD, I really like your software. There is just a little thing to me that can be updated: When using separate processing style, after importing frames when i run "interpolation" if i change the FPS input/output/speed or Interpolation AI the result will be always the same as first selected before "Import/Extract frames". Only the model version can output different results. That make the purpose of "separate processing style" useless since i need to "cleanup" change the settings and rerun "import/extract frames". It would be really good if we were able to change AI, FPS input/output/speed like x2, x3, x4 etc... then rerun the "interpolation" to see different results on the way. This is because for some projects, i need to de-duplicate myself some scenes and to redo this everytime is pretty tiring just to test new results. Also, it would be great to be able to change the output resolution after "import/extract frames" for using a tier software to upscale the imported frames before using "interpolation" because now it show errors when i change the resolution of the pictures. Instead of that, keep up the good work, this software is amazing as it is now :) Just suggesting some improvements !

InoSim

I don't have a log but I only have 1 complaint that started after v1.37.2 of flowframes n is still apparent as of v1.39.0. The acceleration of the interpolation process slows down significantly after 8,000 to 10,000 frames or so using any RIFE version with CUDA n VULKAN both in every version after 1.37.2 with 23fps/29.97fps videos to 240fps. I always use h.265 NVENC GPU accelerated encoding as well. Starts out interpolating with an ETA of 2 or 3 hours, then changes to over 11 hours or more per video after reaching only 10,000 frames. I downgraded to flowframes 1.37.2 again , same settings as above and it shows ETA of 2 hours n 15 min the entire time n currently at frame 104,000 with no slow down. I'm afraid I'll have to keep using v1.37.2 for the speed as I have to interpolate bulks of videos that I need to edit n can't wait 11-15 hours per video. Just wanted to let u know that for future versions of flowframes. I love that v1.37.2 doesn't get any slowdown with interpolation ever n wish that speed stayed for versions after v1.37.2

GamingCanBeFun (YouTuber)

Have you figured out why using de-duplication causes the GPU usage to lower over time? You also can't use NCNN/VS with de-dup, the speed is literally below 0.3 fps

Jake

ayyyy deh wuhh ditit eh cuhh

Kenneth Flournoy

@Nicolas Lietti It must be ur internet connection cuz I was able to download all the models in less than 3 min

GamingCanBeFun (YouTuber)

Same as WaveCut not possible to download models, even the main app took about 30 minutes to download from Germany server and USA server is even slower :S

InoSim

Thank you!

VeryBueno

It's probably firewall related, as windows blocks everything, and you sometimes get an alert to flag that prog, but oftentimes, not! So unless you have a 3rd party prog you'll have to manually trace whatever, silent but the vital program just to flag it as safe!

Kezo

error downloading new models https://i.imgur.com/2Mzk3wv.png

WaveCut

I, for one, am incredibly grateful for having magnificent coders/devs that don't monopolize new tech when developing their software! Thanks again, for your hard, and diligent work with keeping this prog properly updated!

Kezo

Nice, thank you! One thing tho, im getting errors when enabling the option "Make Output Match Input Duration" https://pastebin.com/G8TST8EM

Joel Eriksson

Thank you as well.

fatcat

Damn, straight from 4.1 (Experimental) to 4.4 Just tested it and it seems to have way less artifacts compared to 4.0 Thanks for being a gigachad, making some really good software.

Upscale Anon