Dear Raspberry Pi Team,
My name is Vladislav, and I am developing an autonomous physical assistant based on Raspberry Pi 5 using two Raspberry Pi Camera Module 3 cameras (NoIR, imx708) for stereo vision.
Current configuration:
*Base: Raspberry Pi 5 (16 GB)
*OS: Raspberry Pi OS (Debian Trixie, 64-bit)
*Cameras: 2 × Camera Module 3 NoIR (imx708)
*Connection: CAM0 and CAM1 Pi 5 connectors
*Software: libcamera-apps v1.11.0, rpicam-* utilities
Problem:
An error occurs when trying to launch two cameras simultaneously via rpicam-vid or libcamera.:
terminate called after throwing an instance of 'std::runtime_error'
what(): failed to import fd 53
The analysis shows that the system cannot allocate ISP/PiSP resources for two streams simultaneously, despite the fact that both cameras are correctly detected via rpicam-hello --list-cameras.
Questions:
1. Is simultaneous capture from two Camera Module 3 to Pi 5 supported at the PiSP driver level?
2. If yes, is there a working example (for example, via libcamera API or GStreamer pipeline)?
3. If not, is such support planned in the upcoming updates?
4. Are there any workarounds (for example, interleaving, using different ISP channels)?
Goal:
Implement stereoscopic vision for a local AI assistant with recognition of depth, gaze direction and gestures.
We deeply appreciate your work and hope for your support in this important project for us.
With respect to you and your business.
My name is Vladislav, and I am developing an autonomous physical assistant based on Raspberry Pi 5 using two Raspberry Pi Camera Module 3 cameras (NoIR, imx708) for stereo vision.
Current configuration:
*Base: Raspberry Pi 5 (16 GB)
*OS: Raspberry Pi OS (Debian Trixie, 64-bit)
*Cameras: 2 × Camera Module 3 NoIR (imx708)
*Connection: CAM0 and CAM1 Pi 5 connectors
*Software: libcamera-apps v1.11.0, rpicam-* utilities
Problem:
An error occurs when trying to launch two cameras simultaneously via rpicam-vid or libcamera.:
terminate called after throwing an instance of 'std::runtime_error'
what(): failed to import fd 53
The analysis shows that the system cannot allocate ISP/PiSP resources for two streams simultaneously, despite the fact that both cameras are correctly detected via rpicam-hello --list-cameras.
Questions:
1. Is simultaneous capture from two Camera Module 3 to Pi 5 supported at the PiSP driver level?
2. If yes, is there a working example (for example, via libcamera API or GStreamer pipeline)?
3. If not, is such support planned in the upcoming updates?
4. Are there any workarounds (for example, interleaving, using different ISP channels)?
Goal:
Implement stereoscopic vision for a local AI assistant with recognition of depth, gaze direction and gestures.
We deeply appreciate your work and hope for your support in this important project for us.
With respect to you and your business.
Statistics: Posted by doggerwolf — Tue Jan 20, 2026 5:46 pm