diff --git a/sdk/objc/Framework/Classes/VideoToolbox/RTCVideoEncoderH264.mm b/sdk/objc/Framework/Classes/VideoToolbox/RTCVideoEncoderH264.mm
index f84998efa6..86855a293c 100644
--- a/sdk/objc/Framework/Classes/VideoToolbox/RTCVideoEncoderH264.mm
+++ b/sdk/objc/Framework/Classes/VideoToolbox/RTCVideoEncoderH264.mm
@@ -502,11 +502,10 @@ CFStringRef ExtractProfile(webrtc::SdpVideoFormat videoFormat) {
 
   // If we're capturing native frames in another pixel format than the compression session is
   // configured with, make sure the compression session is reset using the correct pixel format.
+  // If we're capturing non-native frames and the compression session is configured with a non-NV12
+  // format, reset it to NV12.
   OSType framePixelFormat = kNV12PixelFormat;
-  if (pixelBufferPool && [frame.buffer isKindOfClass:[RTCCVPixelBuffer class]]) {
-    RTCCVPixelBuffer *rtcPixelBuffer = (RTCCVPixelBuffer *)frame.buffer;
-    framePixelFormat = CVPixelBufferGetPixelFormatType(rtcPixelBuffer.pixelBuffer);
-
+  if (pixelBufferPool) {
     // The pool attribute `kCVPixelBufferPixelFormatTypeKey` can contain either an array of pixel
     // formats or a single pixel format.
     NSDictionary *poolAttributes =
@@ -520,6 +519,11 @@ CFStringRef ExtractProfile(webrtc::SdpVideoFormat videoFormat) {
       compressionSessionPixelFormats = @[ (NSNumber *)pixelFormats ];
     }
 
+    if ([frame.buffer isKindOfClass:[RTCCVPixelBuffer class]]) {
+      RTCCVPixelBuffer *rtcPixelBuffer = (RTCCVPixelBuffer *)frame.buffer;
+      framePixelFormat = CVPixelBufferGetPixelFormatType(rtcPixelBuffer.pixelBuffer);
+    }
+
     if (![compressionSessionPixelFormats
             containsObject:[NSNumber numberWithLong:framePixelFormat]]) {
       resetCompressionSession = YES;