DAINAPP
patreon
Dain-App 0.39
🕑 Added 2020-05-21 14:17:01 +0000 UTCOk, here is kinda of a big one, let's see:
New "Use Half-Precision float" option
- Make the application use less VRAM by using 16-bit float. It may affect a little the final interpolation and it was related to crash sometimes, but its useful for big resolutions.
Pixel Art tab
- Work in progress tab to interpolate pixel art.
- It already have some options, since for pixel art you don't want some blurry pixels, they should help with that.
New Split workflow
- Now instead of calculating the size of each split, just select the number of divisions you want in the X and Y axis. A lot easier.
UI don't freeze during render
- You can now drag and minimize the UI while the render is going on.
Options
- You can now print the ffmpeg commands to understand more what is going on.
- CRF selection for output
- PNG compression selection (For PNG sequence input and interpolation step)
Bug fixes:
- Pallet used to be created on Dain-APP root. This could make another interpolation to use the pallet from the previous file. Now it's created on the same folder that the other files.
- Fixed PIL timing for gif.
Comments
Not sure about that, but do keep in mind SLI/NVLink doesn't double your VRAM because both GPUs need the same stuff loaded in VRAM to be able to talk to each other properly.
Is "Create a output with audio" working for anyone else?
Then does it support multi-GPU with NVLink? Like 8 V100.
GRisk
The half precision option in 0.39 is the first step to eventually support tensor core, since it need to work with 16fp. The .cuda files still need to be converted to support half precision before starting to experiment with tensor core.
well, when will you add tensor core support? Not every one has Tesla v100.
Sergey Tokarev
Wanted to ask this for some time, but during my experiments I noticed that v0.39 always produce .mp4 output in slowmo (I have to manually increase speed of playback by 2x to see the effect, otherwise playback of output mp4 is slowed down, regardless of framerate). I can provide 0.38 vs 0.39 outputs comparison later, but I'm just curious if I'm missing some option here? With a bit mode testing I found out that Frames Hanling Modes actually works differently in 0.39: Input mp4 (30 fps) outputs: - mode 1: 25 fps file - mode 2: 25 fps file - mode 3: 60 fps file - mode 4: 60 fps file For comparison, 0.38 with mode 1 produced 59.96 fps mp4 file.
