services/viz/privileged/mojom/compositing/frame_sink_video_capture.mojom

// Copyright 2017 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

module viz.mojom;

import "media/capture/mojom/video_capture_buffer.mojom";
import "media/capture/mojom/video_capture_types.mojom";
import "media/mojo/mojom/media_types.mojom";
import "mojo/public/mojom/base/time.mojom";
import "mojo/public/mojom/base/token.mojom";
import "services/viz/public/mojom/compositing/frame_sink_id.mojom";
import "services/viz/public/mojom/compositing/subtree_capture_id.mojom";
import "skia/public/mojom/bitmap.mojom";
import "ui/gfx/geometry/mojom/geometry.mojom";

// Provided with each call to FrameSinkVideoConsumer::OnFrameCaptured() so that
// the consumer can notify the capturer the instant it is done consuming the
// frame, and then later (optionally) provide utilization feedback.
interface FrameSinkVideoConsumerFrameCallbacks {
  // Informs the capturer that the consumer is done reading the content of the
  // video frame.
  Done();

  // |feedback| contains consumer feedback like resource utilization,
  // maximum requested frame-rate and resolution.
  ProvideFeedback(media.mojom.VideoCaptureFeedback feedback);
};

// Provided with FrameSinkVideoConsumer::ChangeTarget() to indicate what
// target should be selected for capture.
union VideoCaptureSubTarget {
  SubtreeCaptureId subtree_capture_id;
  mojo_base.mojom.Token region_capture_crop_id;
};

struct VideoCaptureTarget {
  FrameSinkId frame_sink_id;
  VideoCaptureSubTarget? sub_target;
};

// Interface for a consumer that receives frames and notifications related to
// capture of the source content. An instance that implements this interface is
// provided to FrameSinkVideoCapturer.Start().
interface FrameSinkVideoConsumer {
  // Called to deliver each frame to the consumer. |data| contains the video
  // frame's image data, and is valid for reading until the consumer notifies
  // the service that it is Done(). |info| is used to interpret the
  // format/layout of the data, and also contains the frame timestamps and other
  // metadata (the following media::VideoFrameMetadata keys are set:
  // CAPTURE_BEGIN_TIME, CAPTURE_END_TIME, CAPTURE_COUNTER, CAPTURE_UPDATE_RECT,
  // COLOR_SPACE, FRAME_DURATION, INTERACTIVE_CONTENT, REFERENCE_TIME).
  // |content_rect| is the region of the frame that contains the captured
  // content, with the rest of the frame having been letterboxed to adhere to
  // resolution constraints.
  OnFrameCaptured(media.mojom.VideoBufferHandle data,
                  media.mojom.VideoFrameInfo info,
                  gfx.mojom.Rect content_rect,
                  pending_remote<FrameSinkVideoConsumerFrameCallbacks>
                      callbacks);

  // Called to indicate that a new |sub_capture_target_version| has been set,
  // and that all frames which will subsequently be delivered using
  // OnFrameCaptured(), will have a sub-capture-target-version that is
  // greater-than-or-equal-to this one.
  // This allows the Promise which cropTo() or restrictTo() returned to be
  // resolved promptly, even if frames are not being produced over the track
  // at the moment, which can happen if the track is cropped to an invisible
  // target, or is paused, or is muted.
  OnNewSubCaptureTargetVersion(uint32 sub_capture_target_version);

  // Called when a frame is *not* delivered because Region Capture was applied,
  // but the cropped-to region consists of zero pixels. This is needed because
  // the blue border that is normally drawn around the cropped-to region should
  // be updated.
  // Note that this is specific to Region Capture. With Element Capture,
  // we always draw the blue border around the entire tab's viewport.
  OnFrameWithEmptyRegionCapture();

  // Indicates that OnFrameCaptured() will not be called again, an end-of-stream
  // signal.
  OnStopped();
  // Called to deliver a log message to the consumer.
  // Currently used to populate WebRTC native logs.
  OnLog(string message);
};

// Used to express the preference on the type of VideoBufferHandles that the
// consumer would like to obtain from the capturer. By default, the video
// buffers are returned via read only shared memory.
enum BufferFormatPreference {
  kDefault = 0,
  kPreferGpuMemoryBuffer,
};

// Interface to an implementation that captures the frames of a
// CompositorFrameSink's surface as a video stream. During operation, the
// "target" frame sink can be changed, as needed, to continue capturing the
// desired content (e.g., a new web page in the same browser tab).
//
// The SetXYZ() methods set or change the video capture parameters and
// constraints. They may be called any time before and/or after a Start() call.
// Note that the consumer must not assume any changes have taken effect until
// this is reflected in some later delivered frame.
//
// General usage: A capturer is created, its SetXYZ() methods are called,
// followed by ChangeTarget() and then Start(). Then, later, when video capture
// should suspend, Stop() is called. After that point, Start() can be called
// again to resume capture; or the mojo binding can be dropped to auto-terminate
// the capturer.
interface FrameSinkVideoCapturer {
  // Specifies the pixel format to use. Please note when format is
  // PIXEL_FORMAT_ARGB, the frame might be BGRA or RGBA depends on platform
  // support, please refer to the actual VideoFrameInfo's format at callback.
  //
  // Default, if never called: PIXEL_FORMAT_I420.
  SetFormat(media.mojom.VideoPixelFormat format);

  // Specifies the maximum rate of capture in terms of a minimum time period
  // (min_period = 1/max_frame_rate).
  //
  // Default, if never called: a very conservative 1/5 second (for 5 FPS
  // max). Usually 30-60 FPS is desired for frame sinks that may contain video
  // content.
  SetMinCapturePeriod(mojo_base.mojom.TimeDelta min_period);

  // Specifies the minimum amount of time that must elapse between changing the
  // size of video frames. This can be set to zero to disable resize throttling,
  // which means each video frame could have a different size. Note that
  // setting this period too low can result in erratic frame sizing behaviour;
  // consider disabling auto-throttling in such cases.
  //
  // Default, if never called: 3 seconds.
  SetMinSizeChangePeriod(mojo_base.mojom.TimeDelta min_period);

  // Specifies the range (if any) of acceptable capture resolutions and whether
  // a fixed aspect ratio is required. When |min_size| is equal to |max_size|,
  // capture resolution will be held constant. Otherwise, capture resolution
  // will dynamically adapt over time as the source content size and system
  // resource utilization change. If a fixed aspect ratio is required, the
  // aspect ratio of |max_size| is used and the implementation will letterbox
  // around the captured content when needed.
  //
  // Default, if never called: Fixed at 640x360 (standard definition, 16:9).
  // Usually, 320x180↔3840x2160 (auto-adapt, 16:9) is desired when the consumer
  // can handle variable frame resolutions.
  SetResolutionConstraints(gfx.mojom.Size min_size, gfx.mojom.Size max_size,
                           bool use_fixed_aspect_ratio);

  // Specifies whether capture resolution should be automatically adjusted
  // within the bounds provided to SetReslutionConstraints in response to
  // utilization feedback from the consumer and the buffer pool. When set to
  // false, the capturer will always use the highest resolution within
  // constraints that doesn't exceed the source size.
  //
  // Default, if never called: true.
  SetAutoThrottlingEnabled(bool enabled);

  // Targets a different compositor frame sink. This may be called anytime,
  // before or after Start().
  //
  // If |target| is null, capture will suspend until a new frame sink target
  // is set.
  // If |target| is provided, the frame sink identifier associated with it
  // must be valid.
  //
  // If |target| is a valid subtree capture id, the capturer will
  // capture a render pass associated with a layer subtree under the target
  // frame sink, which is identifiable by that |subtree_capture_id|.
  // Else, if the |target| is a valid sub-capture target, only the
  // relevant sub-capture target will be captured.
  // Otherwise, the capturer captures the root render pass of the target frame
  // sink.
  //
  // |sub_capture_target_version| is a non-decreasing value (for subsequent
  // calls). It is increased whenever the sub-capture target changes
  // (new sub_target). Viz annotates frames with this value, and that allows
  // Blink to discover when a sub-capture has been applied. Blink will resolve
  // a Promise it gave the Web-application when the sub-capture was applied,
  // and this way, the application will know when to start consuming
  // cropped/restricted frames.
  //
  // Note that only tab-capture frames are croppable/restrictable.
  // Anything else, such as window-capture for instance, should always set
  // |sub_capture_target_version| to 0.
  ChangeTarget(VideoCaptureTarget? target, uint32 sub_capture_target_version);

  // Starts emitting video frames to the given |consumer|.
  // The caller can express if the |consumer| prefers to receive video buffers
  // via read only shared memory or via GPU memory buffers by controlling
  // |buffer_format_preference| parameter. If a consumer only supports read only
  // shared memory, `BufferFormatPreference::kDefault` value should be used.
  Start(pending_remote<FrameSinkVideoConsumer> consumer,
        BufferFormatPreference buffer_format_preference);

  // Stops capturing "soon." Meaning, no new frame captures will be started, but
  // there may already be some in-flight for delivery. Wait for the "end of
  // stream" signal provided by FrameSinkVideoConsumer.OnStopped().
  Stop();

  // Requests the capturer send a duplicate of the last frame. This is used to
  // resolve occasional "picture loss" issues consumer-side.
  RequestRefreshFrame();

  // Creates an overlay to be rendered within each captured video frame. The
  // |stacking_index| is an arbitrary value that determines whether to render
  // this overlay before/after other overlays. Greater values mean "after" and
  // "on top of" those with lesser values. Specifying the same index as an
  // existing overlay will cause the existing one to be dropped and replaced
  // with a new one.
  CreateOverlay(int32 stacking_index,
                pending_receiver<FrameSinkVideoCaptureOverlay> receiver);
};

// Control interface for a small image to be composited on top of each captured
// video frame. This allows clients to, for example, have the capturer render
// mouse cursors or debug info boxes on top of the captured content.
interface FrameSinkVideoCaptureOverlay {
  // Sets/Changes the overlay |image| and its position and size, relative to the
  // source content. |bounds| consists of coordinates where the range [0.0,1.0)
  // indicates the relative position+size within the bounds of the source
  // content (e.g., 0.0 refers to the top or left edge; 1.0 to just after the
  // bottom or right edge). Pass empty |bounds| to temporarily hide the overlay
  // until a later call to SetBounds().
  SetImageAndBounds(skia.mojom.BitmapN32 image,
                    gfx.mojom.RectF bounds);

  // Changes the bounds of the previously-set image, showing the overlay if
  // non-empty bounds are provided, and hiding the overlay otherwise. |bounds|
  // has the same semantics as described in ShowImageAt().
  SetBounds(gfx.mojom.RectF bounds);

  // Called to indicate mouse |coordinates| change. |coordinates| are absolute
  // values inside the bound of the source content (e.g., 0 refers to the top
  // or left edge, height-1 to the bottom edge and width-1 to the right edge).
  // The special value (-1, -1) indicates out-of-bound mouse position.
  OnCapturedMouseEvent(gfx.mojom.Point coordinates);
};