Ctrlk
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Changelog

VIESUS version history: what's new, what requires action, and feature additions per release.

Important changes per VIESUS version. Use this to verify which features are available in your installed version, and to plan upgrades.

Upgrading? The configuration file is forward-compatible — load your existing viesusini.json into the latest VIESUS Viewer and save it to add newly available parameters with default values. See FAQ → Can I migrate my existing config? for details.

For the complete PDF Enhancer changelog, see PDF Enhancer Installation.

VIESUS 13

V13 brings updates in scene detection, HDR processing, AI upscaling, facial reconstruction, execution flow, and platform support.

Requires action

  • Enable CPU-based vScene to use scene detection in CPU workflows

  • Set hdrstrength to use HDR-to-SDR tonemapping

  • Turn on VSceneMode=1 to use scene-based 4× AI upscaling

  • Set arstrength to control when Artifact Removal is applied

  • Set VIESUS_UNLOAD if you need GPU model unloading on low-memory systems

Image quality improvements

  • vScene for CPU-based execution — fast scene detection added for CPU workflows. VIESUS detects scene type and applies pre-defined or self-defined parameter settings.

  • HDR — HDR-to-SDR tonemapping with gainmap application for better SDR reproduction. Controlled by hdrstrength.

  • Scene-based 4× AI upscaling — VSceneMode=1 enables scene-based SR. Landscape scenes without people use the sharper 4xAR model.

  • Age group-based Facial Reconstruction — reduces unlikely teeth in baby faces by factoring in age group detection.

  • Improved monochrome AI upscaling — better toned monochrome SR results and RGB facial reconstruction in monochrome detection cases.

Performance improvements

  • Improved 2× AI upscaling — better speed, performance, and consistency.

  • Mixed execution — fast TensorRT for default SR models, CUDA precision for Face Reconstruction.

  • Improved Face Detection — GPU-optimised, no longer blocks the GPU execution pipeline.

  • arstrength parameter — defines the threshold at which Artifact Removal is applied. Skip AR for lightly compressed images that don't justify it.

  • VIESUS_UNLOAD environment variable — unload AI models from the GPU after execution. Trades performance for low-memory GPU support.

Platform support

  • NVIDIA Blackwell GPUs — supported

  • ARM-based Linux systems — supported

VIESUS 12

V12 brings updates in image adjustments, AI upscaling, artifact removal, facial processing, scene and text detection, parameter handling, and GPU infrastructure.

Requires action

  • Use ResizeMode 10 for 2xFast or 4xFast

  • Use ResizeMode 11 for 4xAR

  • Set an Artifact Removal mode if you want the new AR handling

  • Set SDmode = 1 to use panoptic scene detection

  • Set TDmode = 1 to enable experimental text detection

  • Use JSON parameter files if you want JSON-based parameter loading

Image quality improvements

  • Night/Dark adjustment — improved, with shadow correction added for smoother application.

  • AI upscaling — TensorRT improvements for up to 40% performance gain. 2x upscaling improved; new 2xFast model. Additional 4xAR model with integrated Artifact Removal. New modes: ResizeMode 10 for 2xFast/4xFast; ResizeMode 11 for 4xAR.

  • Artifact Removal — improved with multi-mode handling. CPU-based JpegSmoothQuant and AI models combined. Modes: Off=0, OnlyCPU=1, AutoWithAI=2, AutoWithAIFast=3, AlwaysCPU=4, AlwaysAI=5, AlwaysAIFast=6, AlwaysCPUandAI=7, AlwaysCPUandAIFast=8.

  • Face Reconstruction — more reliable for images with multiple faces of different sizes.

  • Background handling — updated models. Greenscreen handling improved.

Detection and analysis

  • Facial Features — facial feature vector added for face similarity and recognition tasks.

  • Scene Detection — SDmode = 1 based on panoptic segmentation.

  • Experimental Text Detection — TDmode = 1 detects areas likely to contain text.

Workflow and platform

  • JSON parameter files — parameter loading now supports JSON.

  • GPU infrastructure — updated to support CUDA 12.6.

VIESUS 11

V11 brings updates in Night/Dark adjustment, AI upscaling, facial processing, background handling, detection, and masking.

Requires action

  • Set Gpars/dastrength to tune Night/Dark adjustment

  • Set srstrength, SRNoiseInjection, IsNoiseRedMode 3, or ISGrainAddMode for new SR and noise controls

  • Use ResizeMode 6, 7, 8, or 9 for new resize modes

  • Use Small Faces, IsFrMode 2/3, or higher fdstrength for new facial options

  • Activate background handling, BGBlur, or AutoCropping where needed

  • Set FDmode = IsFaceDetMode 3 for face anonymisation

Image quality improvements

  • Night/Dark adjustment — configurable adjustments via Gpars/dastrength (default 0.5).

  • Improved AI upscaling — mixed model usage for better SR results.

  • SR blending — srstrength blends AI upscaling with standard upscaling.

  • New resize modes — ResizeMode 6 (no SR + face reconstruction), 7 (2x4x), 8 (2x4xFast), 9 (4xFast).

  • SRNoiseInjection — noise injection before SR.

  • AI Noise Reduction — IsNoiseRedMode 3 removes noise with AI.

  • GrainAddition — ISGrainAddMode adds white Gaussian noise after SR.

Facial processing and detection

  • Small Faces parameter — threshold for model selection; improves SR for small faces.

  • Additional Facial Reconstruction modes — IsFrMode 2 (IS_FR_MIX) and 3 (IS_FR_FAST).

  • Improved Face Detection — primarily for SR. Max faces increased to 512.

  • Face Detection strength — fdstrength > 1.0 (e.g. 2.5) for group photos.

  • Facial Features — eye blink and emotion estimation added.

  • Face anonymisation — IsFaceDetMode 3 blurs all detected faces.

Background handling and cropping

  • Background, Foreground, depth map — introduced.

  • Background Handling — removal to alpha and replacement with image or color.

  • Background Blurring — computational bokeh via Gpars/blurstrength (when BGBlur is active).

  • Background Balancing — introduced.

  • AutoCropping — based on facial parameters.

Masks

  • SDK can provide masks for foreground, background, and depth map.

VIESUS 10

V10 brings updates in AI upscaling, artifact handling, image loading, and face lighting.

Requires action

  • Use AI upscaling as the resize method for the new SR workflow

  • Enable Artifact Detection or Artifact Removal for JPEG cleanup

  • Enable Adaptive Face Flash to brighten dark faces

  • Use the experimental image loader if you need the added format support

Image quality improvements

  • AI upscaling — now available as a separate resizing method (including Face Refinement). Requires NVIDIA GPU.

  • Artifact Detection and Artifact Removal — Detection decides whether to apply removal. Removal reduces JPEG blocking and ringing.

  • Adaptive Face Flash — brightens dark or back-lit faces.

Workflow

  • Experimental Image Loader — loads JPEG, TIFF, PNG, WEBP, HEIC, and RAW formats.

VIESUS 9 and earlier (V9 → V7)

Combined release notes from V9.00 back to V7.00.

V9.00

  • Improved redeye correction — neural network-based with better face detection. False-positive rate dramatically reduced outside facial areas.

  • Natural Skin Enhancement — new parameter to reduce oversaturated, "glowing" faces.

  • White point adjustment for digital printing — light grey pulled to pure white to reduce grey-dot patterns on digital prints.

V8.50

  • Further improved redeye correction — large false positives in eye/nose regions considerably reduced.

  • Noise addition parameter — reduces color banding and posterisation artifacts.

  • Adaptable processing order — redeye reduction can run before all other corrections.

  • iOS library version — direct integration into iOS apps.

V8.00

  • More stable global color and brightness correction — more similar results across similar images.

  • Improved redeye correction — fewer false positives.

  • Smoother redeye strength behavior — continuous variation in detection and false-positive rates.

  • Updated ICC profile handling — LittleCMS 2.4, fully compliant with ICC 4.3.

  • Monochrome conversion — neutral or toned monochrome output (e.g. sepia).

  • Selectable order of resampling and enhancement — opt in to resize-before-enhance.

  • Improved non-sRGB profile handling — direct transform from input to output ICC profile.

V7.50

  • CMYK and grayscale support — previously only RGB.

  • Viewer shadow correction — semi-automatic manual shadow correction in the Viewer.

V7.35

  • Alpha channel support — alpha data used for improved correction.

  • Smaller program footprint — reduced code size.

  • Improved special-case handling — corrupt image data and special ICC profiles handled better.

  • GUID licenses — introduced for special customers.

V7.30

  • New license management — software-based and dongle licenses, plus image-count-based licenses.

  • Folder Enhancer updates — configurable JPEG compression factor, PNG input support.

V7.10

  • Resampling and cropping — added with optimised algorithms for large up/downsampling.

  • Global sharpening parameter — easier device-specific adaptation.

  • Folder Enhancer updates — TIFF input and Unicode file name support.

V7.00

  • Remapped and decoupled enhancement parameters — wider adjustment range, separate brightness and color sliders.

  • Improved redeye correction — fewer false positives at higher detection rate.

  • Continuously adjustable redeye strength.

  • Embedded ICC profile support — embedded profiles in input images taken into account.

  • Output ICC profile support — applied at the end of enhancement.

  • Automatic noise reduction — controllable separately for color and monochrome (off by default).

  • Brighter underexposed images — controllable via global brightness parameter.

PreviousNode.js Error CodesNextVIESUS Assistant

Last updated

Was this helpful?