News, updates, and changes in Composer R1 2025 ("Monaco")
Composer R1 2025
Release date 2025-02-04, version: 2.0.9161
We are proud to present the new AI features of Composer, including an ultrafast Object Detection operator capable of tracking “anything,” such as playing cards, dealers, dice, and more. The operator is compatible with OBB models in the ONNX format. To help you get started, we’ve included a demo tutorial featuring our Blackjack model, trained on thousands of playing card images. The model is available for separate licensing. The Object Detection module in Composer is powered by the latest YOLO SDK from Ultralytics. We have also added a Speech-to-Text operator powered by OpenAI's Whisper technology. Whisper is an automatic speech recognition (ASR) system trained on 680,000 hours of multilingual, multitask supervised data collected from the web.
Our customers have been asking for audio processing filters, and we are excited to announce a new family of operators: Audio Operators. These include a band-pass filter, 10-band and 3-band parametric equalizers, a low-cut filter, Gain, Gate, Delay, a recorder operator, and an audio monitor operator. For more information on audio in Composer, please visit Working with audio in scenes.
We have also updated the audio meter in Composer, adding a detector for overload/distortion and warnings for when the audio level is too high.
Other improvements and new features:
Switcher Input: A new input type that simplifies switching between input sources, with support for cut or crossfade effects.
DeckLink Input: Up to 200% performance improvement, support for selecting a sync source, and compatibility with the DeckLink Mini Recorder series of capture cards.
Project Reference Image: This enables you to store a reference project image and use it to verify the alignment of your physical setup.
Undo Support: Added an undo feature (CTRL+Z) for the most common actions in Composer.
GUI Improvements: Includes a new color picker tool, alignment tools, additional shortcuts, and expanded menu options.
Monitoring and Warnings: Adds support for configuring log retention and other enhancements.
New Monitoring Metrics: Additional metrics have been added to help you monitor the performance of your projects and systems more effectively, ensuring smooth operation.
Bug Fixes: Various bugs and issues reported by users have been addressed, making this version more stable and reliable.
New Tutorials: To help you get the most out of all these new features, new tutorials have been created. These will guide you through the latest capabilities, ensuring you can hit the ground running with Composer R1 2025.
Below is the complete change log:
News and Improvements:
Added AI capabilities to Composer
Added a new operator: Object Detector. This operator supports searching for objects using Onnx models.
A sample Onnx model is included (Coco).
A Blackjack Onnx model is included. This model is a trial version, and it will run for a limited duration (two hours). For a production license, please contact RealSprint.
Added a new operator: Speech to text. This new operator is based on OpenAI Whisper and performs real-time audio-to-text translation (English).
Added support for Audio Operators (Filters):
This version includes the following audio filters: Band-Pass, 10-band Equalizer, 3-band Parametric Eq, Compressor, Low-cut Filter, Gain, Gate, Limiter, Audio Delay, and Audio Monitor.
Added a new input type: Switcher Input. This new input allows you to cut, crossfade or blend between two inputs or scenes.
Decklink Capture: Improved performance - up to 200% performance gain
Video Input: added support for script callback function triggering on End-Of-File
Sync source: added support for using a decklink input as the sync source for the Composer rendering engine. This new feature can improve capture smoothness and reduce dropped- or duplicated frames.
Added configuration for log retention days (default value is 30 days). To change the configuration, edit the settings.xml file in your application folder.
Desktop:
The input preview image of a Rectangle Input or Text Input will automatically update after a setting has been changed.
Rectangle Inputs: the size of newly added rectangles will match the size of the largest scene (width/height in pixels).
The default application lock has been changed from 60 seconds to 600 seconds.
When disabling a layer, the layer compute time will be reset to 0.0ms.
Newly added inputs and scenes will be automatically selected in the list of inputs.
The preview images for Media Inputs, RTMP inputs, and Ultralight inputs will be automatically updated when the input starts.
Added support for aligning layers using the scene preview popup menu.
Added support for moving the anchor point to the center of the layer by using the scene preview popup menu.
Added support for opening project reference images.
Added support for reloading (and restarting) the project script.
Added support reloading (and restarting) the script engine.
Added support for Interactive Mode, which can be used to create clickable objects.
Added support for adding operators using the popup menu for scene layers.
Added support for duplicating inputs.
Added support for overlaying the project reference image in the scene view window.
Added support for automatically renaming all layers to match the name of the source asset/input.
Added support for renaming the selected input to filename only (excluding file path).
Performance section: added tabs in order to fit more information and features.
Added undo support for scene layers (CTRL+Z). The undo feature is currently limited to scene layers.
For all input fields (text, float, double, etc), the Enter key will validate the content and move focus to the next item in the UI.
Double-clicking on a scene name will automatically switch to the inputs tab and select and expand the scene.
A new stereo VU-meter has been added. This new VU-meter will indicate signal overload (above 0db dbfs).
Added support for "expandable sections" in components. This new feature will make it possible to create more compact components without sacrificing functionality.
Added a new color picker tool that supports RGB, HSV, and HSL.
The input preview area is not sizable (vertically), and the preview image will automatically scale to fit the area.
Added warning symbol for high processing time.
Added max processing time (last 50 frames) to GUI.
Added processing queue length to GUI.
Minor GUI overhauls
Performance report:
Added a summary of all assets used by the project.
Added a Settings section, including information on the Lock GPU clocks setting.
Added an option for raising warnings when the audio level averages above 99%.
Decklink capture: added measurements: temperature, PCI link speed and PCI lanes (debug properties)
Improved time-to-first-frame on Windows-based systems.
Script Engine:
Added script support functions:
object GetOperatorObject(string sceneName, string layerName, string operatorName)
string[] GetDirectoryFiles(string path)
Added support for utilizing scripts in inputs and operators.
Added support for reloading of script.
Added support for exporting script to an encrypted script.
Added option for Assets Finder: Search and Replace. This new feature can automatically replace part of a path (filepath or directory path), with an alternative path. Useful when moving projects from one server to another.
Added log retention days configuration. The default value is 30 days, but this can be adjusted in the settings.xml file.
The Blur Operator has been optimized for faster processing. Added support for edge processing options (Border or Wrap).
Added a new BlackmagicOutput (V4) with support for device(s) name and channel index.
Bugfixes:
Settings: Prometheus could sometimes not be enabled, even if a valid license for Extended Monitoring was present.
Scene: the Disable Render option did not silence audio in the scene and did not clear the last rendered frame of the scene.
Media Input: starting media inputs having the "Start playing when loaded" option off, combined with an End-of-file behavior Stop, did not work as expected.
BlackMagic Capture: The single channel capture card BlackMagic Decklink Mini Recorder 4k was incompatible with Composer.
Other changes:
In order for the AI capabilities to work, the following needs to be installed:
CUDA Runtime 11.8
cuDNN 9.4 - CUDA Deep Neural Network)
The Color Channel Keyer has been depricated and removed from the UI. However, projects using the operator will still work.
For new/imported Media Inputs, the NVENC (Hardware accelerated decoding) will be enabled by default.
CUDA version has been updated to 12.4
The setting Enable Performance Log has been removed from the settings window. However, the setting is still available in the settings file.