search

mediar-ai / screenpipe

rewind.ai x cursor.com = AI powered by your 24/7 screen & voice local recording.
https://screenpi.pe
MIT License
8.73k stars 495 forks source link

Windows. screenpipe-vision build without cuda. doesn't work. #393

Closed BAVhel closed 1 month ago

BAVhel commented 1 month ago

screenpipe\target\release>screenpipe-vision.exe -V screenpipe-vision 0.1.93

screenpipe\target\release>screenpipe-vision.exe 2024-10-01T12:00:56.866146Z DEBUG screenpipe_vision::core: continuous_capture: Starting using monitor: 65537 2024-10-01T12:00:56.969284Z ERROR xcap::platform::utils: GetFileVersionInfoSizeW error: WIN32_ERROR(1813) 2024-10-01T12:00:57.433116Z DEBUG screenpipe_vision::core: Captured screenshot on monitor 65537 with hash: 2578426833258961057 2024-10-01T12:00:57.433482Z DEBUG screenpipe_vision::utils: No previous image to compare for frame 0 2024-10-01T12:00:57.456213Z DEBUG screenpipe_vision::core: Performing OCR for frame number since beginning of program 0

2024-10-01T12:00:57.465273Z ERROR screenpipe_vision::core: Error processing OCR task: Unsupported OCR engine 2024-10-01T12:00:58.589977Z ERROR xcap::platform::utils: GetFileVersionInfoSizeW error: WIN32_ERROR(1813) 2024-10-01T12:00:59.065614Z DEBUG screenpipe_vision::core: Captured screenshot on monitor 65537 with hash: 5627007045593965544 2024-10-01T12:00:59.163983Z DEBUG screenpipe_vision::utils: Frame 1: Histogram diff: 0.011, SSIM diff: 0.022, Current Average: 0.016, Max_avr: 0.000 Fr: 0 2024-10-01T12:00:59.189767Z DEBUG screenpipe_vision::core: Performing OCR for frame number since beginning of program 1

linear[bot] commented 1 month ago

MED-132 Windows. screenpipe-vision build without cuda. doesn't work.

louis030195 commented 1 month ago

screenpipe\target\release>screenpipe-vision.exe -V screenpipe-vision 0.1.93

screenpipe\target\release>screenpipe-vision.exe 2024-10-01T12:00:56.866146Z DEBUG screenpipe_vision::core: continuous_capture: Starting using monitor: 65537 2024-10-01T12:00:56.969284Z ERROR xcap::platform::utils: GetFileVersionInfoSizeW error: WIN32_ERROR(1813) 2024-10-01T12:00:57.433116Z DEBUG screenpipe_vision::core: Captured screenshot on monitor 65537 with hash: 2578426833258961057 2024-10-01T12:00:57.433482Z DEBUG screenpipe_vision::utils: No previous image to compare for frame 0 2024-10-01T12:00:57.456213Z DEBUG screenpipe_vision::core: Performing OCR for frame number since beginning of program 0

2024-10-01T12:00:57.465273Z ERROR screenpipe_vision::core: Error processing OCR task: Unsupported OCR engine 2024-10-01T12:00:58.589977Z ERROR xcap::platform::utils: GetFileVersionInfoSizeW error: WIN32_ERROR(1813) 2024-10-01T12:00:59.065614Z DEBUG screenpipe_vision::core: Captured screenshot on monitor 65537 with hash: 5627007045593965544 2024-10-01T12:00:59.163983Z DEBUG screenpipe_vision::utils: Frame 1: Histogram diff: 0.011, SSIM diff: 0.022, Current Average: 0.016, Max_avr: 0.000 Fr: 0 2024-10-01T12:00:59.189767Z DEBUG screenpipe_vision::core: Performing OCR for frame number since beginning of program 1

thx

its kinda hacky code there, i see apple native OCR hardcoded,wont work on windows


    tokio::spawn(async move {
        continuous_capture(
            result_tx,
            Duration::from_secs_f32(1.0 / cli.fps),
            save_text_files,
            OcrEngine::AppleNative,
            id,
            &[],
            &[],
        )
        .await
    });
louis030195 commented 1 month ago

try websocket.rs instead, one thing @m13v did at AGI house hackathon with it:

https://x.com/m13v_/status/1833505959752204368