Selur's Little Message Board - Deoldify Vapoursynth filter

Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146

(04.06.2026, 09:18)hallomanbh Wrote: [ -> ]NOT WORKING ( gray marked)

https://ibb.co/08TbrKJ

Hi hallomanbh,

I developed a GUI for the DiTServerRPC, the main change introduced in HAVC 5.8.5 is the filter CMNet2DiT the same capability has been added to the DiTServerRPC GUI.
You need to carefully follow the instructions for installing the Server and then you need to follow the instructions to install the GUI.

The GUI has the advantage to split in 3 steps the colorization process:

reference frame extraction
reference frame colorization
video encoding using CMNET2

at each step you can check the results before continue in the next step. For example if the reference frames are too few you can lower the threshold to get more, or increase it if are too much. Then you can check the colorized frames, if there is some color inconsistency you can remove the inconsistent frames. Finally you can proceed with the encoding.

Dan

Fresh installed new update Hybrid_dev_2026.06.15-171547 and VapoursynthR77_torch_2026.06.15...i am used RTX 3060 . 12gb.

https://ibb.co/8LZRkY3v

Hi hallomanbh,

you are using the preset speed "slow" which is too slow for your GPU.
in some cases your GPU takes more than 180sec to colorize a frame.
Given that this filter is used inside a vapoursynth pipeline that requires almost real-time encoding there is a timeout of 60sec for request that has been activated in your last frame.

Possible solution:
1) use preset "auto" (should be faster on your GPU)
2) given that you already installed the DitServerRPC, try to install the GUI (that has a connection timeout of 600sec)

Dan

P.S.
I noted from the log that you are using a old version of DitServerRPC, the new version requires cu130 which improves the performances especially of GGUF models, please reinstall it and install also the GUI.

(04.06.2026, 09:42)Selur Wrote: [ -> ]That is correct, and as intended.

Quote: Okay, I will keep CMNet2DiT disabled till it works with R76.
source: https://forum.selur.net/thread-4287-post...l#pid29733

myrsloik has finally released the official R77 with the fix to the deadlock introduced in R76 that was blocking the filter CMNet2DiT to work properly.

Dan

I did release a new version of the torch add-on and dev (yesterday) which come with the official Vapoursynth R77.

Cu Selur

Hello Selur,

but what do you think about this issue: 50% performance regression and new intermittent crashes from R75 to R76

Dan

I did not encounter that and I suspect this should be fixed in R77

Hello Selur,

I was finally able to add DinoV3 in CMNET2 (see : cmnet2) and so I updated also HAVC.
I attached the new RC1 to this post.
The main changes are:

1) new parameter DeepExBackbone, that can have only 2 values: 'dinov2' (default for compatibility), 'dinov3'
2) new weights to be installed for DinoV3:
dinov3-vitb16.zip -> unzip to vshavc\colormnet\weights
DINOv3FeatureV6_LocalAtten_p369412.pth -> download to vshavc\colormnet\weights

Code:
weights/
│   ├── DINOv2FeatureV6_LocalAtten_s2_154000.pth   # ColorMNet pre-trained weights (DINOv2 backbone)
│   ├── DINOv3FeatureV6_LocalAtten_p369412.pth     # Fine-tuned weights (DINOv3 backbone, recommended)
│   └── dinov3-vitb16/                             # DINOv3 ViT-B/16 backbone (HuggingFace format)

in the GUI, the new parameter DeepExBackbone must be enabled only if is selected the "Exemplar Model" CMNET2

I hope that you have time to include this version in Hybrid.

Thanks,
Dan

P.S.
The training of the model is still a work in progress, but the released model weights are already quite good, see the images in this folder for a fair comparison

Probably won't get around to test this before Saturday, but the comparison images look nice. Smile

Speed is probabl still horrible, on my hardware, but it's nice to see progress in the results.
(mainly working on Linux and MacOS stuff atm. and originally planned to look into Vapoursynth R80 at the weekend)

Cu Selur

Hi Dan, Hi Selur,
That’s really great news! I’m very happy that you were able to integrate DINOv3 into CMNET2. DINOv3 seems significantly more powerful than DINOv2, so I’m really looking forward to testing it.
I hope it can also be adapted to Hybrid for the colorization workflow, especially for the ReColor option. That could be a very interesting improvement.
Thank you very much for your work and for continuing to improve HAVC and Hybrid!
Best
NASS

@Dan64: using:

Code:
clip = havc.HAVC_main(clip, DeepExMethod=3, ScFrameDir=None, DeepExModel=0, DeepExEncMode=0, DeepExBackbone="dinov3", DeepExMaxMemFrames=20)

I get:

Code:
2026-09-19 15:03:03.073
Failed to evaluate the script:
Python exception: CopyFrameProps: argument prop_src is required

Traceback (most recent call last):
File "vapoursynth.pyx", line 3832, in vapoursynth._vpy_evaluate
File "vapoursynth.pyx", line 3833, in vapoursynth._vpy_evaluate
File "J:\tmp\tempPreviewVapoursynthFile15_02_58_577.vpy", line 39, in 
clip = havc.HAVC_main(clip, DeepExMethod=3, ScFrameDir=None, DeepExModel=0, DeepExEncMode=0, DeepExBackbone="dinov3", DeepExMaxMemFrames=20)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "f:\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\__init__.py", line 325, in HAVC_main
return HAVC_main_presets(clip, Preset, FrameInterp, ColorModel, CombMethod, VideoTune, ColorFix,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "f:\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\__init__.py", line 512, in HAVC_main_presets
clip_colored = HAVC_main_colorizer(clip, Preset, ColorModel, CombMethod, VideoTune, ColorFix, ColorTemp,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "f:\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\__init__.py", line 874, in HAVC_main_colorizer
clip_colored = HAVC_cmnet2(clip, clip_ref=None, method=DeepExMethod, render_speed=DeepExPreset,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "f:\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\__init__.py", line 2001, in HAVC_cmnet2
clip = CopySCDetect(clip, clip_ref)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "f:\hybrid\64bit\vapoursynth\Lib\site-packages\vshavc\vsslib\vsscdect.py", line 159, in CopySCDetect
return clip.std.CopyFrameProps(prop_src=sc, props=['_SceneChangePrev', '_SceneChangeNext',
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "vapoursynth.pyx", line 3559, in vapoursynth.Function.__call__
vapoursynth.Error: CopyFrameProps: argument prop_src is required

Cu Selur

Ps.: updated Hybrid_havc_test