diff --git a/screenpipe-vision/Cargo.toml b/screenpipe-vision/Cargo.toml
index 92792dd1c..c9f2b8f7a 100644
--- a/screenpipe-vision/Cargo.toml
+++ b/screenpipe-vision/Cargo.toml
@@ -43,6 +43,9 @@ tracing = { workspace = true }
 which = "6.0"
 serde = "1.0.200"
 
+# TO-DO: use a released version. Done for now to avoid waiting xcap release.
+xcap = { package = "xcap", git = "https://github.com/nashaofu/xcap", rev = "214cb44" }
+
 once_cell = { workspace = true }
 base64 = "0.22.1"
 
@@ -96,13 +99,10 @@ windows = { version = "0.58", features = [
   "Storage",
   "Storage_Streams",
 ] }
-xcap = "0.0.12"
 
 [target.'cfg(target_os = "macos")'.dependencies]
 libc = "=0.2.164"
-xcap-macos = { package = "xcap", git = "https://github.com/mediar-ai/xcap", rev = "965bc99" }
 cidre = { git = "https://github.com/yury/cidre.git", version = "0.5.0" }
 
 [target.'cfg(target_os = "linux")'.dependencies]
 libc = "=0.2.164"
-xcap = "0.0.14"
diff --git a/screenpipe-vision/src/capture_screenshot_by_window.rs b/screenpipe-vision/src/capture_screenshot_by_window.rs
index a87248219..4524ce7b3 100644
--- a/screenpipe-vision/src/capture_screenshot_by_window.rs
+++ b/screenpipe-vision/src/capture_screenshot_by_window.rs
@@ -7,10 +7,6 @@ use std::fmt;
 use std::time::Duration;
 use tokio::time;
 
-#[cfg(target_os = "macos")]
-use xcap_macos::{Monitor, Window, XCapError};
-
-#[cfg(not(target_os = "macos"))]
 use xcap::{Monitor, Window, XCapError};
 
 #[derive(Debug)]
@@ -238,7 +234,7 @@ pub async fn capture_all_visible_windows(
                     image,
                     app_name: app_name.to_string(),
                     window_name: window_name.to_string(),
-                    is_focused: is_valid,
+                    is_focused: window.is_focused(),
                 });
             }
             Err(e) => error!(
@@ -260,13 +256,8 @@ pub fn is_valid_window(
     capture_unfocused_windows: bool,
 ) -> bool {
     if !capture_unfocused_windows {
-        // Early returns for simple checks
-        #[cfg(target_os = "macos")]
         let is_focused = window.current_monitor().id() == monitor.id() && window.is_focused();
 
-        #[cfg(not(target_os = "macos"))]
-        let is_focused = window.current_monitor().id() == monitor.id() && !window.is_minimized();
-
         if !is_focused {
             return false;
         }
diff --git a/screenpipe-vision/src/core.rs b/screenpipe-vision/src/core.rs
index b45b70b15..bb992143d 100644
--- a/screenpipe-vision/src/core.rs
+++ b/screenpipe-vision/src/core.rs
@@ -33,10 +33,6 @@ use tokio::sync::watch;
 use tracing::info;
 use tracing::warn;
 
-#[cfg(target_os = "macos")]
-use xcap_macos::Monitor;
-
-#[cfg(not(target_os = "macos"))]
 use xcap::Monitor;
 
 fn serialize_image<S>(image: &Option<DynamicImage>, serializer: S) -> Result<S::Ok, S::Error>
diff --git a/screenpipe-vision/src/monitor.rs b/screenpipe-vision/src/monitor.rs
index f96196751..133a0889a 100644
--- a/screenpipe-vision/src/monitor.rs
+++ b/screenpipe-vision/src/monitor.rs
@@ -1,10 +1,5 @@
-#[cfg(target_os = "macos")]
-use xcap_macos::Monitor;
-
-#[cfg(not(target_os = "macos"))]
 use xcap::Monitor;
 
-
 pub async fn list_monitors() -> Vec<Monitor> {
     Monitor::all().unwrap().to_vec()
 }
diff --git a/screenpipe-vision/src/utils.rs b/screenpipe-vision/src/utils.rs
index 28a8f2349..9a21585a7 100644
--- a/screenpipe-vision/src/utils.rs
+++ b/screenpipe-vision/src/utils.rs
@@ -9,10 +9,6 @@ use log::{debug, error, warn};
 use std::hash::{DefaultHasher, Hash, Hasher};
 use std::time::{Duration, Instant};
 
-#[cfg(target_os = "macos")]
-use xcap_macos::Monitor;
-
-#[cfg(not(target_os = "macos"))]
 use xcap::Monitor;
 
 #[derive(Clone, Debug, Default)]