FAQ
Brief answers about the features, data and limitations of the current version of Screph.
What is Screph for?
Screph links visual sources to markup, relationships, descriptions, and computer vision results. The saved data can be used when working with code or a separate automation script.
Does Screph start automation automatically after annotation?
No. The project and script are saved separately, and Automation Runtime is run as a separate command after checking the target and input method.
Does a processing result immediately change the project?
No. The mask, outline, detection, or other result is first shown separately. A supported change to the underlying markup is applied only after user confirmation.
Which modes are ready?
General and GUI Automation support the core workflows. Industrial and UAV provide experimental domain tools; each action has its own availability checks. Complete production control is not available, and Ground Robotics remains deferred.
Do you need an account?
To work locally with sources, markup, projects and local computer vision, you do not need an account. It is used for profile, support and available server operations.
What data can leave the computer?
Text, audio, images, or context are sent when the user selects a cloud operation, submits support, or submits diagnostics. Checking for updates and loading models transfers the usual technical network request data, but not the project and frames.
Is it possible to work without the Internet?
Core local workflows work without the cloud. Local speech recognition requires an installed model. A local LLM requires a configured environment or service for the selected connection. Cloud functions, updates, and downloads of missing models require a network connection.
Do you need a separate video card?
No, it is not required for the main program. NVIDIA-supported CUDA can speed up individual methods with models if the environment and model are configured appropriately.
Where are projects stored?
In the directory selected when saving. The project can reference images and results next to it, so you need to copy the entire project directory to transfer.
Download Voice2Text
The Voice2Text page shows current download availability and file details.
Can network operations be purchased?
No. The website does not currently have published rates or a working purchase process.
Where can I see the current status?
The product status page lists mode and component status and download availability.
Why is a result marked stale after I change settings?
A result belongs to the data and settings used when the run started. After changing them, prepare the current input and run processing again. The previous run remains available in history.
How is a processing chain different from a frame sequence?
A chain connects processing steps. A sequence describes frames over time. In Frame sequences, choose a source range and available frame processing; adding a step to a chain does not by itself create temporal analysis.
Do I need a training environment to annotate a dataset?
Manual annotation and standard dataset export do not require a training environment. Training and model-dependent operations do. Before a run, the selected task, model, device, and required resources are checked.
What happens to training when I close the window?
This depends on the selected launch mode. A detached background run can continue after the interface closes; an attached run ends with its application owner. To cancel, use the stop command and wait for confirmation. After a failure, recovery or resume is available only for a suitable saved state.
Why is the primary model absent from my final program?
Promoting and selecting a primary model records the training outcome. For the final program, separately define its application, input source, and output use, then explicitly include the model in the delivered materials.
Does camera calibration immediately provide sizes and distances?
Calibration describes the camera and helps correct image geometry. Measurements in physical units require suitable observations and a spatial reference. Check the profile’s applicability and validate results using independent observations.
Is ROS installed with Screph?
No. Preparing materials for a code agent works without ROS. Data exchange and continuous processing require a separately configured ROS Python environment and an explicit connection. The integration handles images and perception results; it does not turn the deferred Ground Robotics mode into a ready robot-control system.
Does in-app help include every website page?
Interactive mode opens the website’s entire help catalog inside the application, including learning examples. It requires a network connection. Saved mode provides the index and five core guides offline. If the website is unavailable, the application switches to saved help. A failed update of pages or illustrations does not remove the previous working set.