Dear Topaz Team
To further enhance and improve the software, I would like to submit the following suggestions, which I believe would significantly contribute to its functionality and user experience.
1. Intelligent Video Scaling Based on the Longer Side:
Automatic Scaling by the Longer Side: I propose introducing an option for automatic video scaling based on the longer side. The software should automatically detect the longer side of the video (whether horizontal or vertical) and apply scaling to the desired resolution based on that dimension. This would ensure that the video is scaled correctly regardless of its orientation, similar to the functionality found in programs like Photo AI and Gigapixel. Essentially, this is a classic “scale to fit” option, but automated and adapted to video orientation.
2. AI Model and Performance Enhancements:
TensorRT Implementation for AION Model: I recommend implementing TensorRT optimization for the AION model, which would significantly speed up the inference process and improve software performance.
AI Models for Lighting, White Balance, and Contrast: It would be extremely beneficial to integrate AI models similar to “adjust lighting v2” from the upcoming Photo AI for automatic adjustment of lighting and white balance (considering that a white balance model already exists and is usable). Additionally, I propose introducing options for precise contrast adjustment, giving users greater control over the final video appearance.
Specialized AI Models:
Artifact Removal Model: Development and training of an AI model specifically designed for the efficient removal of various video artifacts.
VHS Material Model: Creation of a model specifically trained to optimize upscaling and restoration of video material recorded on VHS tapes.
Audio Upscaling Model: Development of an AI model for upscaling the audio components of video files to 5.1 and DTS formats, thereby improving audio quality as well.
3. Consistent Performance Across Model Versions:
Maintain Performance of Older Models: It would be highly desirable if previous versions of AI models, particularly older Proteus models, could operate at the same processing speed as newer versions. Ensuring consistent performance across different model iterations would improve user workflow and predictability.
4. Video Analysis Tools:
Professional Video Scopes: Integration of video analysis tools such as parade, waveform, vectorscope, and histogram would enable more precise control and correction of color, exposure, and other video parameters.
5. Audio Processing and VST Plugin Support:
VST Plugin Wrapper: Implementing a VST plugin wrapper would allow users to apply a wide range of audio effects to the video sound using third-party VST plugins. This would enable advanced audio processing such as automatic loudness leveling, equalization, dialogue extraction, plosive removal, and more.
6. Upscaling Quality Optimization:
Blurriness Issue at 4x Upscaling: It has been observed that with certain models, better image quality is achieved in 1x mode compared to 4x upscaling, especially with lower resolution input videos. It is necessary to optimize model performance at higher scaling factors, especially for lower resolution inputs, to avoid blurriness and ensure consistent quality.
7. User Interface Personalization:
Option to Revert to Old Skins: Providing an option to revert to skins from previous major versions of the software would offer users greater flexibility and the ability to personalize the user interface.
Thank you for considering these suggestions. If I think of any further suggestions, I will gladly share them.
Sincerely,