diff --git a/src/gfx_api.rs b/src/gfx_api.rs
index 018ba13b..2af944fc 100644
--- a/src/gfx_api.rs
+++ b/src/gfx_api.rs
@@ -265,11 +265,23 @@ pub trait GfxFramebuffer: Debug {
         clear: Option<&Color>,
     ) -> Result<Option<SyncFile>, GfxError>;
 
-    fn copy_to_shm(self: Rc<Self>, shm: &[Cell<u8>]) -> Result<(), GfxError>;
-
     fn format(&self) -> &'static Format;
 }
 
+pub trait GfxInternalFramebuffer: GfxFramebuffer {
+    fn into_fb(self: Rc<Self>) -> Rc<dyn GfxFramebuffer>;
+
+    fn staging_size(&self) -> usize;
+
+    fn download(
+        self: Rc<Self>,
+        staging: &Rc<dyn GfxStagingBuffer>,
+        callback: Rc<dyn AsyncShmGfxTextureCallback>,
+        mem: Rc<dyn ShmMemory>,
+        damage: Region,
+    ) -> Result<Option<PendingShmTransfer>, GfxError>;
+}
+
 impl dyn GfxFramebuffer {
     pub fn clear(
         &self,
@@ -593,13 +605,14 @@ pub trait GfxContext: Debug {
 
     fn gfx_api(&self) -> GfxApi;
 
-    fn create_fb(
+    fn create_internal_fb(
         self: Rc<Self>,
+        cpu_worker: &Rc<CpuWorker>,
         width: i32,
         height: i32,
         stride: i32,
         format: &'static Format,
-    ) -> Result<Rc<dyn GfxFramebuffer>, GfxError>;
+    ) -> Result<Rc<dyn GfxInternalFramebuffer>, GfxError>;
 
     fn sync_obj_ctx(&self) -> Option<&Rc<SyncObjCtx>>;
 
diff --git a/src/gfx_apis/gl/renderer/context.rs b/src/gfx_apis/gl/renderer/context.rs
index e0ab21bb..32cf2392 100644
--- a/src/gfx_apis/gl/renderer/context.rs
+++ b/src/gfx_apis/gl/renderer/context.rs
@@ -5,7 +5,7 @@ use {
         format::{Format, XRGB8888},
         gfx_api::{
             AsyncShmGfxTexture, BufferResvUser, GfxContext, GfxError, GfxFormat, GfxFramebuffer,
-            GfxImage, ResetStatus, ShmGfxTexture,
+            GfxImage, GfxInternalFramebuffer, ResetStatus, ShmGfxTexture,
         },
         gfx_apis::gl::{
             egl::{context::EglContext, display::EglDisplay, image::EglImage},
@@ -316,13 +316,14 @@ impl GfxContext for GlRenderContext {
         GfxApi::OpenGl
     }
 
-    fn create_fb(
+    fn create_internal_fb(
         self: Rc<Self>,
+        _cpu_worker: &Rc<CpuWorker>,
         width: i32,
         height: i32,
         _stride: i32,
         format: &'static Format,
-    ) -> Result<Rc<dyn GfxFramebuffer>, GfxError> {
+    ) -> Result<Rc<dyn GfxInternalFramebuffer>, GfxError> {
         let fb = self.ctx.with_current(|| unsafe {
             GlRenderBuffer::new(&self.ctx, width, height, format)?.create_framebuffer()
         })?;
diff --git a/src/gfx_apis/gl/renderer/framebuffer.rs b/src/gfx_apis/gl/renderer/framebuffer.rs
index 071f3a29..bbe98c16 100644
--- a/src/gfx_apis/gl/renderer/framebuffer.rs
+++ b/src/gfx_apis/gl/renderer/framebuffer.rs
@@ -1,7 +1,11 @@
 use {
     crate::{
         format::Format,
-        gfx_api::{AcquireSync, GfxApiOpt, GfxError, GfxFramebuffer, ReleaseSync, SyncFile},
+        gfx_api::{
+            AcquireSync, AsyncShmGfxTextureCallback, GfxApiOpt, GfxError, GfxFramebuffer,
+            GfxInternalFramebuffer, GfxStagingBuffer, PendingShmTransfer, ReleaseSync, ShmMemory,
+            SyncFile,
+        },
         gfx_apis::gl::{
             gl::{
                 frame_buffer::GlFrameBuffer,
@@ -13,6 +17,7 @@ use {
             sys::{GL_ONE, GL_ONE_MINUS_SRC_ALPHA},
             RenderError,
         },
+        rect::Region,
         theme::Color,
     },
     std::{
@@ -104,11 +109,31 @@ impl GfxFramebuffer for Framebuffer {
         self.render(acquire_sync, ops, clear).map_err(|e| e.into())
     }
 
-    fn copy_to_shm(self: Rc<Self>, shm: &[Cell<u8>]) -> Result<(), GfxError> {
-        (*self).copy_to_shm(shm).map_err(|e| e.into())
-    }
-
     fn format(&self) -> &'static Format {
         self.gl.rb.format
     }
 }
+
+impl GfxInternalFramebuffer for Framebuffer {
+    fn into_fb(self: Rc<Self>) -> Rc<dyn GfxFramebuffer> {
+        self
+    }
+
+    fn staging_size(&self) -> usize {
+        0
+    }
+
+    fn download(
+        self: Rc<Self>,
+        _staging: &Rc<dyn GfxStagingBuffer>,
+        _callback: Rc<dyn AsyncShmGfxTextureCallback>,
+        mem: Rc<dyn ShmMemory>,
+        _damage: Region,
+    ) -> Result<Option<PendingShmTransfer>, GfxError> {
+        let mut res = Ok(());
+        mem.access(&mut |mem| res = self.copy_to_shm(mem))
+            .map_err(RenderError::AccessFailed)?;
+        res?;
+        Ok(None)
+    }
+}
diff --git a/src/gfx_apis/vulkan.rs b/src/gfx_apis/vulkan.rs
index 1193acd9..14e2ca82 100644
--- a/src/gfx_apis/vulkan.rs
+++ b/src/gfx_apis/vulkan.rs
@@ -23,7 +23,7 @@ use {
         cpu_worker::{jobs::read_write::ReadWriteJobError, CpuWorker},
         format::Format,
         gfx_api::{
-            AsyncShmGfxTexture, GfxContext, GfxError, GfxFormat, GfxFramebuffer, GfxImage,
+            AsyncShmGfxTexture, GfxContext, GfxError, GfxFormat, GfxImage, GfxInternalFramebuffer,
             GfxStagingBuffer, ResetStatus, ShmGfxTexture, StagingBufferUsecase, STAGING_DOWNLOAD,
             STAGING_UPLOAD,
         },
@@ -316,16 +316,23 @@ impl GfxContext for Context {
         GfxApi::Vulkan
     }
 
-    fn create_fb(
+    fn create_internal_fb(
         self: Rc<Self>,
+        cpu_worker: &Rc<CpuWorker>,
         width: i32,
         height: i32,
         stride: i32,
         format: &'static Format,
-    ) -> Result<Rc<dyn GfxFramebuffer>, GfxError> {
-        let fb = self
-            .0
-            .create_shm_texture(format, width, height, stride, &[], true, None)?;
+    ) -> Result<Rc<dyn GfxInternalFramebuffer>, GfxError> {
+        let fb = self.0.create_shm_texture(
+            format,
+            width,
+            height,
+            stride,
+            &[],
+            true,
+            Some(cpu_worker),
+        )?;
         Ok(fb)
     }
 
diff --git a/src/gfx_apis/vulkan/image.rs b/src/gfx_apis/vulkan/image.rs
index b005c7bc..43fcb376 100644
--- a/src/gfx_apis/vulkan/image.rs
+++ b/src/gfx_apis/vulkan/image.rs
@@ -4,8 +4,8 @@ use {
         gfx_api::{
             AcquireSync, AsyncShmGfxTexture, AsyncShmGfxTextureCallback,
             AsyncShmGfxTextureTransferCancellable, GfxApiOpt, GfxError, GfxFramebuffer, GfxImage,
-            GfxStagingBuffer, GfxTexture, PendingShmTransfer, ReleaseSync, ShmGfxTexture,
-            ShmMemory, SyncFile,
+            GfxInternalFramebuffer, GfxStagingBuffer, GfxTexture, PendingShmTransfer, ReleaseSync,
+            ShmGfxTexture, ShmMemory, SyncFile,
         },
         gfx_apis::vulkan::{
             allocator::VulkanAllocation, device::VulkanDevice, format::VulkanModifierLimits,
@@ -510,17 +510,46 @@ impl GfxFramebuffer for VulkanImage {
             .map_err(|e| e.into())
     }
 
-    fn copy_to_shm(self: Rc<Self>, shm: &[Cell<u8>]) -> Result<(), GfxError> {
-        self.renderer
-            .read_all_pixels(&self, shm)
-            .map_err(|e| e.into())
-    }
-
     fn format(&self) -> &'static Format {
         self.format
     }
 }
 
+impl GfxInternalFramebuffer for VulkanImage {
+    fn into_fb(self: Rc<Self>) -> Rc<dyn GfxFramebuffer> {
+        self
+    }
+
+    fn staging_size(&self) -> usize {
+        let VulkanImageMemory::Internal(shm) = &self.ty else {
+            unreachable!();
+        };
+        shm.size as _
+    }
+
+    fn download(
+        self: Rc<Self>,
+        staging: &Rc<dyn GfxStagingBuffer>,
+        callback: Rc<dyn AsyncShmGfxTextureCallback>,
+        mem: Rc<dyn ShmMemory>,
+        damage: Region,
+    ) -> Result<Option<PendingShmTransfer>, GfxError> {
+        let VulkanImageMemory::Internal(shm) = &self.ty else {
+            unreachable!();
+        };
+        let staging = staging.clone().into_vk(&self.renderer.device.device);
+        let pending = shm.async_transfer(
+            &self,
+            staging,
+            &mem,
+            damage,
+            callback,
+            TransferType::Download,
+        )?;
+        Ok(pending)
+    }
+}
+
 impl GfxTexture for VulkanImage {
     fn size(&self) -> (i32, i32) {
         (self.width as _, self.height as _)
diff --git a/src/gfx_apis/vulkan/renderer.rs b/src/gfx_apis/vulkan/renderer.rs
index e3feb025..053877ad 100644
--- a/src/gfx_apis/vulkan/renderer.rs
+++ b/src/gfx_apis/vulkan/renderer.rs
@@ -32,11 +32,10 @@ use {
     ash::{
         vk,
         vk::{
-            AccessFlags2, AttachmentLoadOp, AttachmentStoreOp, BufferImageCopy,
-            BufferMemoryBarrier2, ClearColorValue, ClearValue, CommandBuffer,
-            CommandBufferBeginInfo, CommandBufferSubmitInfo, CommandBufferUsageFlags,
-            CopyImageInfo2, DependencyInfo, DependencyInfoKHR, DescriptorImageInfo, DescriptorType,
-            Extent2D, Extent3D, Fence, ImageAspectFlags, ImageCopy2, ImageLayout,
+            AccessFlags2, AttachmentLoadOp, AttachmentStoreOp, ClearColorValue, ClearValue,
+            CommandBuffer, CommandBufferBeginInfo, CommandBufferSubmitInfo,
+            CommandBufferUsageFlags, CopyImageInfo2, DependencyInfoKHR, DescriptorImageInfo,
+            DescriptorType, Extent2D, Extent3D, ImageAspectFlags, ImageCopy2, ImageLayout,
             ImageMemoryBarrier2, ImageSubresourceLayers, ImageSubresourceRange, PipelineBindPoint,
             PipelineStageFlags2, Rect2D, RenderingAttachmentInfo, RenderingInfo,
             SemaphoreSubmitInfo, SemaphoreSubmitInfoKHR, ShaderStageFlags, SubmitInfo2, Viewport,
@@ -49,7 +48,7 @@ use {
     std::{
         cell::{Cell, RefCell},
         fmt::{Debug, Formatter},
-        mem, ptr,
+        mem,
         rc::Rc,
         slice,
     },
@@ -905,143 +904,6 @@ impl VulkanRenderer {
         frame.waiter.set(Some(future));
     }
 
-    pub(super) fn read_all_pixels(
-        self: &Rc<Self>,
-        tex: &VulkanImage,
-        dst: &[Cell<u8>],
-    ) -> Result<(), VulkanError> {
-        let Some(shm_info) = &tex.format.shm_info else {
-            return Err(VulkanError::UnsupportedShmFormat(tex.format.name));
-        };
-        let size = tex.stride as u64 * tex.height as u64;
-        if size != dst.len() as u64 {
-            return Err(VulkanError::InvalidBufferSize);
-        }
-        let region = BufferImageCopy::default()
-            .buffer_row_length(tex.stride / shm_info.bpp)
-            .buffer_image_height(tex.height)
-            .image_subresource(ImageSubresourceLayers {
-                aspect_mask: ImageAspectFlags::COLOR,
-                mip_level: 0,
-                base_array_layer: 0,
-                layer_count: 1,
-            })
-            .image_extent(Extent3D {
-                width: tex.width,
-                height: tex.height,
-                depth: 1,
-            });
-        let staging =
-            self.device
-                .create_staging_buffer(&self.allocator, size, false, true, true)?;
-        let initial_tex_barrier;
-        let initial_buffer_barrier = BufferMemoryBarrier2::default()
-            .buffer(staging.buffer)
-            .offset(0)
-            .size(staging.size)
-            .dst_access_mask(AccessFlags2::TRANSFER_WRITE)
-            .dst_stage_mask(PipelineStageFlags2::TRANSFER);
-        let mut initial_barriers = DependencyInfo::default()
-            .buffer_memory_barriers(slice::from_ref(&initial_buffer_barrier));
-        if tex.bridge.is_none() {
-            initial_tex_barrier = image_barrier()
-                .src_queue_family_index(QUEUE_FAMILY_FOREIGN_EXT)
-                .dst_queue_family_index(self.device.graphics_queue_idx)
-                .image(tex.image)
-                .old_layout(ImageLayout::GENERAL)
-                .new_layout(ImageLayout::TRANSFER_SRC_OPTIMAL)
-                .dst_access_mask(AccessFlags2::TRANSFER_READ)
-                .dst_stage_mask(PipelineStageFlags2::TRANSFER);
-            initial_barriers =
-                initial_barriers.image_memory_barriers(slice::from_ref(&initial_tex_barrier));
-        }
-        let final_tex_barrier;
-        let final_buffer_barrier = BufferMemoryBarrier2::default()
-            .buffer(staging.buffer)
-            .offset(0)
-            .size(staging.size)
-            .src_access_mask(AccessFlags2::TRANSFER_WRITE)
-            .src_stage_mask(PipelineStageFlags2::TRANSFER)
-            .dst_access_mask(AccessFlags2::HOST_READ)
-            .dst_stage_mask(PipelineStageFlags2::HOST);
-        let mut final_barriers = DependencyInfo::default()
-            .buffer_memory_barriers(slice::from_ref(&final_buffer_barrier));
-        if tex.bridge.is_none() {
-            final_tex_barrier = image_barrier()
-                .src_queue_family_index(self.device.graphics_queue_idx)
-                .dst_queue_family_index(QUEUE_FAMILY_FOREIGN_EXT)
-                .image(tex.image)
-                .old_layout(ImageLayout::TRANSFER_SRC_OPTIMAL)
-                .new_layout(ImageLayout::GENERAL)
-                .src_access_mask(AccessFlags2::TRANSFER_READ)
-                .src_stage_mask(PipelineStageFlags2::TRANSFER);
-            final_barriers =
-                final_barriers.image_memory_barriers(slice::from_ref(&final_tex_barrier));
-        }
-        let buf = self.gfx_command_buffers.allocate()?;
-        let mut semaphores = vec![];
-        let mut semaphore_infos = vec![];
-        if let VulkanImageMemory::DmaBuf(buf) = &tex.ty {
-            for plane in &buf.template.dmabuf.planes {
-                let fd = dma_buf_export_sync_file(&plane.fd, DMA_BUF_SYNC_READ)
-                    .map_err(VulkanError::IoctlExportSyncFile)?;
-                let semaphore = self.allocate_semaphore()?;
-                semaphore.import_sync_file(fd)?;
-                let semaphore_info = SemaphoreSubmitInfo::default()
-                    .semaphore(semaphore.semaphore)
-                    .stage_mask(PipelineStageFlags2::TOP_OF_PIPE);
-                semaphores.push(semaphore);
-                semaphore_infos.push(semaphore_info);
-            }
-        }
-        let command_buffer_info = CommandBufferSubmitInfo::default().command_buffer(buf.buffer);
-        let submit_info = SubmitInfo2::default()
-            .wait_semaphore_infos(&semaphore_infos)
-            .command_buffer_infos(slice::from_ref(&command_buffer_info));
-        let begin_info =
-            CommandBufferBeginInfo::default().flags(CommandBufferUsageFlags::ONE_TIME_SUBMIT);
-        unsafe {
-            self.device
-                .device
-                .begin_command_buffer(buf.buffer, &begin_info)
-                .map_err(VulkanError::BeginCommandBuffer)?;
-            self.device
-                .device
-                .cmd_pipeline_barrier2(buf.buffer, &initial_barriers);
-            self.device.device.cmd_copy_image_to_buffer(
-                buf.buffer,
-                tex.image,
-                ImageLayout::TRANSFER_SRC_OPTIMAL,
-                staging.buffer,
-                &[region],
-            );
-            self.device
-                .device
-                .cmd_pipeline_barrier2(buf.buffer, &final_barriers);
-            self.device
-                .device
-                .end_command_buffer(buf.buffer)
-                .map_err(VulkanError::EndCommandBuffer)?;
-            self.device
-                .device
-                .queue_submit2(
-                    self.device.graphics_queue,
-                    slice::from_ref(&submit_info),
-                    Fence::null(),
-                )
-                .map_err(VulkanError::Submit)?;
-        }
-        self.block();
-        self.gfx_command_buffers.buffers.push(buf);
-        for semaphore in semaphores {
-            self.wait_semaphores.push(semaphore);
-        }
-        staging.download(|mem, size| unsafe {
-            ptr::copy_nonoverlapping(mem, dst.as_ptr() as _, size);
-        })?;
-        Ok(())
-    }
-
     pub fn execute(
         self: &Rc<Self>,
         fb: &VulkanImage,
diff --git a/src/ifs/zwlr_screencopy_frame_v1.rs b/src/ifs/zwlr_screencopy_frame_v1.rs
index 44afe8c5..41183bd8 100644
--- a/src/ifs/zwlr_screencopy_frame_v1.rs
+++ b/src/ifs/zwlr_screencopy_frame_v1.rs
@@ -2,6 +2,7 @@ use {
     crate::{
         client::{Client, ClientError},
         format::XRGB8888,
+        gfx_api::{AsyncShmGfxTextureCallback, GfxError, PendingShmTransfer},
         ifs::{
             wl_buffer::{WlBuffer, WlBufferError, WlBufferStorage},
             wl_output::OutputGlobalOpt,
@@ -9,6 +10,7 @@ use {
         leaks::Tracker,
         object::{Object, Version},
         rect::Rect,
+        utils::errorfmt::ErrorFmt,
         wire::{zwlr_screencopy_frame_v1::*, WlBufferId, ZwlrScreencopyFrameV1Id},
     },
     std::{cell::Cell, ops::Deref, rc::Rc},
@@ -29,6 +31,7 @@ pub struct ZwlrScreencopyFrameV1 {
     pub with_damage: Cell<bool>,
     pub buffer: Cell<Option<Rc<WlBuffer>>>,
     pub version: Version,
+    pub pending: Cell<Option<PendingShmTransfer>>,
 }
 
 impl ZwlrScreencopyFrameV1 {
@@ -132,6 +135,7 @@ impl ZwlrScreencopyFrameV1 {
             node.screencopies.remove(&(self.client.id, self.id));
             node.screencast_changed();
         }
+        self.pending.take();
     }
 }
 
@@ -153,6 +157,22 @@ impl ZwlrScreencopyFrameV1RequestHandler for ZwlrScreencopyFrameV1 {
     }
 }
 
+impl AsyncShmGfxTextureCallback for ZwlrScreencopyFrameV1 {
+    fn completed(self: Rc<Self>, res: Result<(), GfxError>) {
+        self.pending.take();
+        match res {
+            Ok(_) => {
+                let now = self.client.state.now();
+                self.send_ready(now.0.tv_sec as _, now.0.tv_nsec as _);
+            }
+            Err(e) => {
+                log::warn!("Could not perform shm screencopy: {}", ErrorFmt(e));
+                self.send_failed();
+            }
+        }
+    }
+}
+
 object_base! {
     self = ZwlrScreencopyFrameV1;
     version = self.version;
diff --git a/src/ifs/zwlr_screencopy_manager_v1.rs b/src/ifs/zwlr_screencopy_manager_v1.rs
index ae99a4b4..66520274 100644
--- a/src/ifs/zwlr_screencopy_manager_v1.rs
+++ b/src/ifs/zwlr_screencopy_manager_v1.rs
@@ -130,6 +130,7 @@ impl ZwlrScreencopyManagerV1 {
             with_damage: Cell::new(false),
             buffer: Cell::new(None),
             version: self.version,
+            pending: Default::default(),
         });
         track!(self.client, frame);
         self.client.add_client_obj(&frame)?;
diff --git a/src/it/test_gfx_api.rs b/src/it/test_gfx_api.rs
index e6a0de03..87d33a81 100644
--- a/src/it/test_gfx_api.rs
+++ b/src/it/test_gfx_api.rs
@@ -6,8 +6,8 @@ use {
         gfx_api::{
             AcquireSync, AsyncShmGfxTexture, AsyncShmGfxTextureCallback, CopyTexture, FillRect,
             FramebufferRect, GfxApiOpt, GfxContext, GfxError, GfxFormat, GfxFramebuffer, GfxImage,
-            GfxStagingBuffer, GfxTexture, GfxWriteModifier, PendingShmTransfer, ReleaseSync,
-            ResetStatus, ShmGfxTexture, ShmMemory, SyncFile,
+            GfxInternalFramebuffer, GfxStagingBuffer, GfxTexture, GfxWriteModifier,
+            PendingShmTransfer, ReleaseSync, ResetStatus, ShmGfxTexture, ShmMemory, SyncFile,
         },
         rect::{Rect, Region},
         theme::Color,
@@ -165,13 +165,14 @@ impl GfxContext for TestGfxCtx {
         GfxApi::OpenGl
     }
 
-    fn create_fb(
+    fn create_internal_fb(
         self: Rc<Self>,
+        _cpu_worker: &Rc<CpuWorker>,
         width: i32,
         height: i32,
         stride: i32,
         format: &'static Format,
-    ) -> Result<Rc<dyn GfxFramebuffer>, GfxError> {
+    ) -> Result<Rc<dyn GfxInternalFramebuffer>, GfxError> {
         assert!(stride >= width * 4);
         Ok(Rc::new(TestGfxFb {
             img: Rc::new(TestGfxImage::Shm(TestShmGfxImage {
@@ -548,15 +549,34 @@ impl GfxFramebuffer for TestGfxFb {
         Ok(None)
     }
 
-    fn copy_to_shm(self: Rc<Self>, shm: &[Cell<u8>]) -> Result<(), GfxError> {
-        self.img.deref().read_pixels(shm)
-    }
-
     fn format(&self) -> &'static Format {
         &ARGB8888
     }
 }
 
+impl GfxInternalFramebuffer for TestGfxFb {
+    fn into_fb(self: Rc<Self>) -> Rc<dyn GfxFramebuffer> {
+        self
+    }
+
+    fn staging_size(&self) -> usize {
+        0
+    }
+
+    fn download(
+        self: Rc<Self>,
+        _staging: &Rc<dyn GfxStagingBuffer>,
+        _callback: Rc<dyn AsyncShmGfxTextureCallback>,
+        mem: Rc<dyn ShmMemory>,
+        _damage: Region,
+    ) -> Result<Option<PendingShmTransfer>, GfxError> {
+        let mut res = Ok(());
+        mem.access(&mut |mem| res = self.img.deref().read_pixels(mem))
+            .map_err(TestGfxError::AccessFailed)?;
+        res.map(|_| None)
+    }
+}
+
 impl dyn GfxTexture {
     fn as_native(&self) -> &TestGfxImage {
         self.as_any()
diff --git a/src/state.rs b/src/state.rs
index 75a13882..f9222f65 100644
--- a/src/state.rs
+++ b/src/state.rs
@@ -27,8 +27,8 @@ use {
         forker::ForkerProxy,
         format::Format,
         gfx_api::{
-            AcquireSync, BufferResv, GfxContext, GfxError, GfxFramebuffer, GfxTexture, ReleaseSync,
-            SampleRect, SyncFile,
+            AcquireSync, BufferResv, GfxContext, GfxError, GfxFramebuffer, GfxTexture,
+            PendingShmTransfer, ReleaseSync, SampleRect, SyncFile, STAGING_DOWNLOAD,
         },
         gfx_apis::create_gfx_context,
         globals::{Globals, GlobalsError, RemovableWaylandGlobal, WaylandGlobal},
@@ -62,7 +62,7 @@ use {
         io_uring::IoUring,
         leaks::Tracker,
         logger::Logger,
-        rect::Rect,
+        rect::{Rect, Region},
         renderer::Renderer,
         scale::Scale,
         security_context_acceptor::SecurityContextAcceptors,
@@ -1014,25 +1014,32 @@ impl State {
         x_off: i32,
         y_off: i32,
         size: Option<(i32, i32)>,
-        capture: &ZwlrScreencopyFrameV1,
-        mem: &ClientMemOffset,
+        capture: &Rc<ZwlrScreencopyFrameV1>,
+        mem: &Rc<ClientMemOffset>,
         stride: i32,
         format: &'static Format,
         transform: Transform,
         scale: Scale,
-    ) -> Result<(), ShmScreencopyError> {
+    ) -> Result<Option<PendingShmTransfer>, ShmScreencopyError> {
         let Some(ctx) = self.render_ctx.get() else {
             return Err(ShmScreencopyError::NoRenderContext);
         };
         let fb = ctx
-            .create_fb(capture.rect.width(), capture.rect.height(), stride, format)
+            .clone()
+            .create_internal_fb(
+                &self.cpu_worker,
+                capture.rect.width(),
+                capture.rect.height(),
+                stride,
+                format,
+            )
             .map_err(ShmScreencopyError::CreateTemporaryFb)?;
         self.perform_screencopy(
             src,
             None,
             acquire_sync,
             ReleaseSync::None,
-            &fb,
+            &fb.clone().into_fb(),
             AcquireSync::Unnecessary,
             ReleaseSync::None,
             transform,
@@ -1045,14 +1052,16 @@ impl State {
             scale,
         )
         .map_err(ShmScreencopyError::CopyToTemporary)?;
-        let acc = mem.access(|mem| fb.copy_to_shm(mem));
-        match acc {
-            Ok(res) => res.map_err(ShmScreencopyError::ReadPixels),
-            Err(e) => {
-                capture.client.error(e);
-                Ok(())
-            }
-        }
+        let staging = ctx.create_staging_buffer(fb.staging_size(), STAGING_DOWNLOAD);
+        let pending = fb
+            .download(
+                &staging,
+                capture.clone(),
+                mem.clone(),
+                Region::new2(capture.rect.at_point(0, 0)),
+            )
+            .map_err(ShmScreencopyError::ReadPixels)?;
+        Ok(pending)
     }
 
     pub fn create_seat(self: &Rc<Self>, name: &str) -> Rc<WlSeatGlobal> {
diff --git a/src/tree/output.rs b/src/tree/output.rs
index d0dffce6..92f85598 100644
--- a/src/tree/output.rs
+++ b/src/tree/output.rs
@@ -267,6 +267,7 @@ impl OutputNode {
                 capture.send_failed();
                 continue;
             }
+            let mut ready = true;
             if let Some(storage) = wl_buffer.storage.borrow_mut().deref() {
                 match storage {
                     WlBufferStorage::Shm { mem, stride } => {
@@ -284,10 +285,16 @@ impl OutputNode {
                             self.global.persistent.transform.get(),
                             self.global.persistent.scale.get(),
                         );
-                        if let Err(e) = res {
-                            log::warn!("Could not perform shm screencopy: {}", ErrorFmt(e));
-                            capture.send_failed();
-                            continue;
+                        match res {
+                            Ok(p) => {
+                                ready = p.is_none();
+                                capture.pending.set(p);
+                            }
+                            Err(e) => {
+                                log::warn!("Could not perform shm screencopy: {}", ErrorFmt(e));
+                                capture.send_failed();
+                                continue;
+                            }
                         }
                     }
                     WlBufferStorage::Dmabuf { fb, .. } => {
@@ -327,7 +334,9 @@ impl OutputNode {
             if capture.with_damage.get() {
                 capture.send_damage();
             }
-            capture.send_ready(now.0.tv_sec as _, now.0.tv_nsec as _);
+            if ready {
+                capture.send_ready(now.0.tv_sec as _, now.0.tv_nsec as _);
+            }
         }
         self.screencast_changed();
     }