New! Highlight threshold slider. This slider explicitly tells the model the lowest intensity level to be considered a highlight. This slider should be set to a value higher than the brightest skin-tone (or any other color where accuracy is vital) present in the video. In practical terms, we recommend starting with the default setting and raising the value if undesirable color shifts appear.
New! Saturation boost slider. The HDR color space has a larger color gamut and supports more saturated colors. Turning this slider up results in more saturated outputs in HDR without loss of detail. This slider will, however, lead to a color shift.
Rhea XL (Previously RXL) Alpha 3
Speed improvement on 30 and 40 series Nvidia GPUs due to TensorRT conversion.
Optimized for more input sizes.
Known Issues
Exporting Hyperion results with H265 Main10 on macOS can create color shifts over time. We are not aware of any other encoder or setting with this issue.
We would like to clarify how the exposure slider in Hyperion affects the HDR output’s brightness. By design, the model always produces HDR videos with a peak brightness of 1000 nits, regardless of the exposure setting. Importantly, the model always generates HDR videos with a peak brightness of 1000 nits, regardless of the exposure setting. The slider’s primary function is to adjust the brightness of shadows and mid-tones.
With the exposure value set to 0, the shadows and mid-tones in the HDR video should closely match those of the input SDR video when viewed on a 200 nit monitor, disregarding the wider color gamut. The highlights will be expanded to fill the remaining part of the intensity range.
Increasing the slider to 1 brightens the shadows and mid-tones by one exposure stop, doubling their brightness compared to the 0 setting. This adjustment allows for fine-tuning of the overall image brightness while maintaining the 1000 nits peak brightness.
As always, we welcome feedback and questions as we get closer to launch. We thank you for your continued participation in the alphas!
Interesting additional info, I will definitely use it to refine the brightness and hue of Hyperion’s work.
It’s not that simple actually.
/i like the zoom of the timeline with the borders if I use the magnifying glass
/i don’t always like a mega zoom on the first mouse wheel turn
Rhea XL still has some (huge) problems with static shapes with a moving background. It generates a huge blur around the edge
ex: edge of an airplane wing in front of a scrolling landscape
EDIT : Issue with (.mov (prores) ) RheaXL, ffprobe reporting error with codec, impossible to read or set avg_frame_rate, video stream is choppy
→ Workaround (.mkv (prores)) fixed the problem
The preview playback inside the application is still completely broken, exiting the application still doesn’t remove it from the memory (I had to kill it manually with System Informer),
Frame Interpolation doesn’t seem to work (at least with original frame rate and duplicate frames at 10%; it seems to work a bit for camera panning with conversion to 60fps [from 29.97], but the cartoon animations themselves aren’t interpolated at all).
When the app is ‘hung’ with the playback (can’t stop it nor seek the video anymore), the preview processing will arbitrarily start from frame 0, no matter where the marker is placed. I’m terribly sorry to ask that, but were the VEAI 5.3.4.4b beta and this alpha tested with the basic functionality?
The Rhea XL model is promising though!
Ah, I’d forget - the Live Render is somehow connected to the preview playback - also doesn’t seem to work in this alpha - playback in this mode just does nothing, not even a model is being downloaded.
Trial Mode persists despite validating the app twice (it functions correctly though, without watermark, cause the auth.tpz file is being written correctly).
If only it would be possible to patch VEAI 5.0.4 to support Gaia v6, Rhea and RheaXL… Oh, right, why is the support of using additional models (like Gaia v6 or Rhea for v5.0.4) by placing their corresponding JSON files in models folder removed? Granted the models are of the same ‘architecture’, they should work, probably, right?
Oh well, Rhea XL is good but it has a huge problem with backgrounds, because everything that moves in the foreground is smeared with the background, individual objects that are in the foreground are always smeared with the moving background
Rhea XL yields really nice results with 720p and higher input resolutions. At least, with my inputs
For lower resolution inputs, I get visually more pleasing results when I use a 2x FemaSR ONNX model in Hybrid (with the vs-mlrt addon and fp16 TensorRT support) as a first pass.
I upscale the input video 2x in Hybrid, and then process the output file with the Rhea XL model. It’s visually more appealing and sharper, but sometimes there are discrepancies where text becomes illegible and such.
It’s definitely worth trying different combinations to see what works best. And different onnx models if FemaSR not work for your input well. FeMaSR_SRX2_model_g .onnx link Hybrid addons (vs-mlrt) Hybrid installer
Thanks to the Topaz team for the continuous development of the program and the effort to fix all potential problems and bugs.
What Tab in Hybrid you have those settings (and possibility to load the AI onnx)?
I can’t seem to find it.
I have the Mod addons by Selur installed as I use DeOldify coloring that needs those.
Topaz Video AI has proven to be quite effective and fast for my needs. I’ve noticed a significant improvement with the Rhea XL model compared to the standard Rhea model, especially when processing higher-resolution footage or using it in a second pass workflow (whether through Topaz Video AI itself, DaVinci Resolve’s upscaling, or other methods – the possibilities are numerous). Ultimately, everyone should choose the workflow that best suits their needs.
Frankly, I don’t understand the need for irony and sarcasm, and I believe such a tone is inappropriate for forum discussions.
This log is only intended for user information. It should not be part of a bug/problem report!!
Hybrid: [homepage](http://www.selur.de)
Forum: [public forum](http://forum.selur.net)
Regarding problems please read: [needed infos](http://www.selur.de/support)
Donate: [via PayPal](https://www.paypal.com/cgi-bin/webscr?cmd=_s-xclick&hosted_button_id=4H65FUWQTCRFN)
This log is only intended for user information. It should not be part of a bug/problem report!!
Detected NVIDIA PureVideo compatible cards: NVIDIA GeForce GTX 1080 Ti
Detected vfwDecoders with 32bit: VIDC.LAGS VIDC.X264 VIDC.XVID vidc.cvid vidc.i420 vidc.iyuv vidc.mrle vidc.msvc vidc.uyvy vidc.yuy2 vidc.yvu9 vidc.yvyu
Detected vfw64BitDecoders: VIDC.LAGS VIDC.X264 VIDC.XVID vidc.i420 vidc.iyuv vidc.mrle vidc.msvc vidc.uyvy vidc.yuy2 vidc.yvu9 vidc.yvyu
Encounter XML error reading 'readGeneral': Opening and ending tag mismatch.
Avisynth+ is available,..
DGDecNV available,..
Vapoursynth is available,..
DGDecNV available,..
VSGAN available,..
vsDPIR available,..
vsRIFE (torch) available,..
vsBasicVSR++ available,..
vsRealESRGAN available,..
vsSwinIR available,..
vsHINet available,..
vsAnimeSR available,..
vsFeMaSR available,..
vsSCUNet available,..
vsCodeFormer available,..
vsGRLIR available,..
vsDDColor available,..
vsDeOldify available,..
vsProPainter available,..
vsDeepDeinterlace available,..
VSMLRT available,..
vsDPIR (mlrt) available,..
vsRIFE (mlrt) available,..
vsSAFA (mlrt) available,..
vsSCUNet mlrt) available,..
vsSwinIR (mlrt) available,..
Resetting Vapoursynth filter order. Adjust&Save the new FilteringModel defaults.
Resetting Avisynth filter order. Adjust/Rest your FilteringModel defaults.
Current defaultOutputPath: is empty!
-> Set it under Config->Path
Finished startup, finished after 9.661s
I really don’t know. Did you perhaps accidentally extract it into a subfolder? The file structure should look like this.
Resizing filters are under Vapoursynth section–Frame-Resize
It’s best to ask Selur on his forum.
link link
You could try, for example, ScuNET alongside traditional upscaling methods (to your desired resolution) in Hybrid as a “first pass” before using Rhea XL. Interestingly, Hybrid can be “tuned” to blend, meaning you can adjust the strength of the effect (you need to enable the “show -merge controls” option).
