From d8ea2afe477b7f8bc995f010506f7faec34a7ebd Mon Sep 17 00:00:00 2001
From: Calixte Denizet <calixte.denizet@gmail.com>
Date: Tue, 23 Jun 2026 12:04:09 +0200
Subject: [PATCH] Add support for Screen annotations playing embedded media

Screen annotations whose rendition action resolves to an embedded audio/video
file now play through the same play-button overlay as RichMedia.
Factor the shared resolution logic into a MediaAnnotation base (used by both RichMedia and Screen).

It fixes #6078 and #2787.
---
 src/core/annotation.js                      | 347 +++++++++++++++----
 src/core/file_spec.js                       |  10 +
 src/display/annotation_layer.js             |  15 +-
 src/shared/util.js                          |  11 +
 test/integration/annotation_spec.mjs        |  63 +++-
 test/integration/presentation_mode_spec.mjs |  11 +-
 test/unit/annotation_spec.js                | 365 +++++++++++++++++++-
 web/annotation_layer_builder.css            |  10 +-
 web/annotation_layer_builder.js             |   2 +-
 web/pdf_presentation_mode.js                |   6 +-
 web/viewer.css                              |   2 +-
 11 files changed, 737 insertions(+), 105 deletions(-)

diff --git a/src/core/annotation.js b/src/core/annotation.js
index 6761df712..8eed43b5e 100644
--- a/src/core/annotation.js
+++ b/src/core/annotation.js
@@ -19,6 +19,7 @@ import {
   AnnotationEditorType,
   AnnotationFieldFlag,
   AnnotationFlag,
+  AnnotationRenditionOperation,
   AnnotationReplyType,
   AnnotationType,
   assert,
@@ -290,6 +291,9 @@ class AnnotationFactory {
       case "RichMedia":
         return new RichMediaAnnotation(parameters);
 
+      case "Screen":
+        return new ScreenAnnotation(parameters);
+
       default:
         if (!collectFields) {
           if (!subtype) {
@@ -5457,13 +5461,124 @@ class FileAttachmentAnnotation extends MarkupAnnotation {
   }
 }
 
-class RichMediaAnnotation extends Annotation {
+/**
+ * Shared base for annotations that play an embedded audio/video clip:
+ * `RichMedia` (via `RichMediaContent`) and `Screen` (via a rendition action).
+ * Both resolve a single embedded media file and expose it identically through
+ * `data.richMedia`, so the display layer can render them with one element.
+ */
+class MediaAnnotation extends Annotation {
+  // The MIME types we can build a `<video>`/`<audio>` element for.
+  static #MEDIA_MIME_TYPE_RE = /^(?:video|audio)\//;
+
   constructor(params) {
     super(params);
 
+    // No HTML element until a playable asset is found below by the subclass.
     this.data.noHTML = true;
+  }
+
+  /**
+   * Expose a resolved embedded media asset as `data.richMedia`.
+   *
+   * @param {Object} asset
+   * @param {Ref | null} asset.assetRef
+   *   Reference to the file-spec dictionary (or, for an inline file-spec, the
+   *   embedded-file stream); used to lazily fetch the bytes on the main thread.
+   * @param {Dict} asset.assetDict
+   *   The file-spec (or stream) dictionary, used to locate the embedded file
+   *   when `assetRef` isn't itself a reference.
+   * @param {string} asset.filename
+   * @param {string} asset.contentType
+   * @param {Catalog} [catalog]
+   */
+  _setMediaData({ assetRef, assetDict, filename, contentType }, catalog) {
+    let contentRef = assetRef;
+    if (!(contentRef instanceof Ref)) {
+      contentRef = FileSpec.pickPlatformItem(
+        assetDict.get("EF"),
+        /* raw = */ true
+      );
+    }
+    const fileId =
+      contentRef instanceof Ref
+        ? catalog?.getAttachmentIdForAnnotation(contentRef)
+        : undefined;
+
+    this.data.noHTML = false;
+    this.data.richMedia = { fileId, filename, contentType };
+  }
+
+  /**
+   * Determine the MIME type used to build the `<video>`/`<audio>` element.
+   *
+   * When the media dictionary provides an explicit type (e.g. a MediaClip
+   * `/CT`), it's honored if it names an audio/video type. Otherwise, per the
+   * spec (ISO 32000-2, 7.11.4) an embedded file stream declares its MIME type
+   * through its own `/Subtype`, with characters not allowed in a name
+   * hex-escaped (e.g. `/video#2Fmp4` -> `video/mp4`). We trust that when it
+   * names an audio/video type, and otherwise fall back to mapping the filename
+   * extension. Returns `null` when the asset isn't a recognized audio/video
+   * type (e.g. Flash `.swf` or 3D models), so we don't build a player that
+   * can't play anything.
+   *
+   * @param {Dict} assetDict
+   * @param {string} filename
+   * @param {string | null} [contentType]
+   * @returns {string | null}
+   */
+  static _getContentType(assetDict, filename, contentType = null) {
+    if (
+      typeof contentType === "string" &&
+      MediaAnnotation.#MEDIA_MIME_TYPE_RE.test(contentType)
+    ) {
+      return contentType;
+    }
+
+    // The embedded file stream is keyed like a file-spec platform item.
+    const stream = FileSpec.pickPlatformItem(assetDict.get("EF"));
+    const subtype =
+      stream instanceof BaseStream ? stream.dict?.get("Subtype") : null;
+    if (
+      subtype instanceof Name &&
+      MediaAnnotation.#MEDIA_MIME_TYPE_RE.test(subtype.name)
+    ) {
+      return subtype.name;
+    }
+
+    const ext = filename.split(".").at(-1)?.toLowerCase();
+    switch (ext) {
+      case "mp4":
+      case "m4v":
+        return "video/mp4";
+      case "webm":
+        return "video/webm";
+      case "ogv":
+        return "video/ogg";
+      case "mov":
+        return "video/quicktime";
+      case "mp3":
+        return "audio/mpeg";
+      case "m4a":
+        return "audio/mp4";
+      case "wav":
+        return "audio/wav";
+      case "oga":
+      case "ogg":
+        return "audio/ogg";
+      default:
+        return null;
+    }
+  }
+}
+
+class RichMediaAnnotation extends MediaAnnotation {
+  constructor(params) {
+    super(params);
 
     const { dict, xref, annotationGlobals } = params;
+    /** @type {{catalog?: Catalog}} */
+    const { catalog } = annotationGlobals.pdfManager.pdfDocument;
 
     const content = dict.get("RichMediaContent");
     if (!(content instanceof Dict)) {
@@ -5475,24 +5590,8 @@ class RichMediaAnnotation extends Annotation {
       warn("RichMedia annotation has no playable asset.");
       return;
     }
-    const { assetRef, assetDict, filename, contentType } = asset;
 
-    let contentRef = assetRef;
-    if (!(contentRef instanceof Ref)) {
-      contentRef = FileSpec.pickPlatformItem(
-        assetDict.get("EF"),
-        /* raw = */ true
-      );
-    }
-    const fileId =
-      contentRef instanceof Ref
-        ? annotationGlobals.pdfManager.pdfDocument.catalog?.getAttachmentIdForAnnotation(
-            contentRef
-          )
-        : undefined;
-
-    this.data.noHTML = false;
-    this.data.richMedia = { fileId, filename, contentType };
+    this._setMediaData(asset, catalog);
   }
 
   /**
@@ -5543,11 +5642,12 @@ class RichMediaAnnotation extends Annotation {
         if (!(asset instanceof Dict)) {
           continue;
         }
+        // Skip assets that only reference an external file we can't read.
+        if (!FileSpec.hasEmbeddedFile(asset)) {
+          continue;
+        }
         const { filename } = new FileSpec(asset).serializable;
-        const contentType = RichMediaAnnotation.#getContentType(
-          asset,
-          filename
-        );
+        const contentType = MediaAnnotation._getContentType(asset, filename);
         if (!contentType) {
           continue;
         }
@@ -5562,54 +5662,177 @@ class RichMediaAnnotation extends Annotation {
 
     return null;
   }
+}
+
+class ScreenAnnotation extends MediaAnnotation {
+  constructor(params) {
+    super(params);
+
+    const { dict, xref, annotationGlobals } = params;
+    const asset = ScreenAnnotation.#findAsset(dict, xref);
+    if (!asset) {
+      // Not every Screen annotation plays embedded media (e.g. a URL stream or
+      // a /Movie); such ones simply render their appearance, so don't warn.
+      return;
+    }
+    this._setMediaData(asset, annotationGlobals.pdfManager.pdfDocument.catalog);
+  }
 
   /**
-   * Determine the MIME type used to build the `<video>`/`<audio>` element.
+   * Locate the embedded media played by the annotation's rendition action.
    *
-   * Per the spec (ISO 32000-2, 7.11.4) an embedded file stream declares its
-   * MIME type through its own `/Subtype`, with characters not allowed in a
-   * name hex-escaped (e.g. `/video#2Fmp4` -> `video/mp4`). We trust that when
-   * it names an audio/video type, and otherwise fall back to mapping the
-   * filename extension. Returns `null` when the asset isn't a recognized
-   * audio/video type (e.g. Flash `.swf` or 3D models), so we don't build a
-   * player that can't play anything.
+   * Per the spec (ISO 32000-1, 12.6.4.13 and 13.2) the chain is:
+   *   Screen `/A` (or `/AA`) rendition action -> `/R` rendition (`/MR`)
+   *   -> `/C` media clip (`/MCD`) -> `/D` file-spec -> `/EF` embedded file.
+   * Selector renditions (`/SR`) are unwrapped to their first playable media
+   * rendition. This mirrors `RichMediaAnnotation`, which also targets the
+   * common single embedded-media case.
    *
-   * @param {Dict} assetDict
-   * @param {string} filename
-   * @returns {string | null}
+   * @returns {{
+   *   assetRef: Ref | null,
+   *   assetDict: Dict,
+   *   filename: string,
+   *   contentType: string,
+   * } | null}
    */
-  static #getContentType(assetDict, filename) {
-    // The embedded file stream is keyed like a file-spec platform item.
-    const stream = FileSpec.pickPlatformItem(assetDict.get("EF"));
-    const subtype =
-      stream instanceof BaseStream ? stream.dict?.get("Subtype") : null;
-    if (subtype instanceof Name && /^(?:video|audio)\//.test(subtype.name)) {
-      return subtype.name;
+  static #findAsset(dict, xref) {
+    for (const action of this.#renditionActions(dict, xref)) {
+      const asset = this.#findRenditionAsset(
+        action.get("R"),
+        xref,
+        new RefSet()
+      );
+      if (asset) {
+        return asset;
+      }
+    }
+    return null;
+  }
+
+  static *#renditionActions(dict, xref) {
+    // The rendition action may be the activation action (/A) or one of the
+    // additional actions (/AA), e.g. page-open.
+    const action = xref.fetchIfRef(dict.getRaw("A"));
+    if (
+      action instanceof Dict &&
+      isName(action.get("S"), "Rendition") &&
+      this.#isPlayAction(action)
+    ) {
+      yield action;
+    }
+    const additionalActions = dict.get("AA");
+    if (additionalActions instanceof Dict) {
+      for (const key of additionalActions.getKeys()) {
+        const aa = xref.fetchIfRef(additionalActions.getRaw(key));
+        if (
+          aa instanceof Dict &&
+          isName(aa.get("S"), "Rendition") &&
+          this.#isPlayAction(aa)
+        ) {
+          yield aa;
+        }
+      }
+    }
+  }
+
+  static #isPlayAction(action) {
+    // Rendition action /OP (ISO 32000-1, Table 214): PLAY_OR_RESUME and PLAY
+    // play; STOP/PAUSE/RESUME don't start playback. When absent, the action is
+    // JS-driven (/JS), which we can't run, so assume play.
+    const operation = action.get("OP");
+    return (
+      operation === undefined ||
+      operation === AnnotationRenditionOperation.PLAY_OR_RESUME ||
+      operation === AnnotationRenditionOperation.PLAY
+    );
+  }
+
+  static #findRenditionAsset(rendition, xref, seen) {
+    if (!(rendition instanceof Dict)) {
+      return null;
+    }
+    const subtype = rendition.get("S");
+    if (isName(subtype, "MR")) {
+      return this.#findClipAsset(rendition.get("C"), xref);
+    }
+    if (isName(subtype, "SR")) {
+      // A selector rendition lists candidate renditions; play the first that
+      // resolves to embedded media.
+      const renditions = rendition.get("R");
+      if (Array.isArray(renditions)) {
+        for (const ref of renditions) {
+          // Guard against renditions referencing each other in a cycle.
+          if (ref instanceof Ref) {
+            if (seen.has(ref)) {
+              continue;
+            }
+            seen.put(ref);
+          }
+          const asset = this.#findRenditionAsset(
+            xref.fetchIfRef(ref),
+            xref,
+            seen
+          );
+          if (asset) {
+            return asset;
+          }
+        }
+      }
+    }
+    return null;
+  }
+
+  static #findClipAsset(clip, xref) {
+    if (!(clip instanceof Dict) || !isName(clip.get("S"), "MCD")) {
+      return null;
+    }
+    const rawData = clip.getRaw("D");
+    const data = xref.fetchIfRef(rawData);
+    const contentTypeHint = clip.get("CT");
+    let explicitType =
+      typeof contentTypeHint === "string" ? contentTypeHint : null;
+
+    let assetDict, filename;
+    if (data instanceof BaseStream) {
+      // `/D` is the embedded media stream directly.
+      assetDict = data.dict;
+      // `/N` is a human-readable label, not a filename, so it's an unreliable
+      // source for a file extension. When `/CT` is absent, prefer the stream's
+      // own `/Subtype` if it declares a media MIME type (as embedded-file
+      // streams do); `_getContentType` ignores a non-media value.
+      const name = clip.get("N");
+      filename = typeof name === "string" ? stringToPDFString(name) : "";
+      if (!explicitType) {
+        const subtype = data.dict.get("Subtype");
+        if (subtype instanceof Name) {
+          explicitType = subtype.name;
+        }
+      }
+    } else if (data instanceof Dict) {
+      // `/D` is a file specification; require a readable embedded file.
+      if (!FileSpec.hasEmbeddedFile(data)) {
+        return null;
+      }
+      assetDict = data;
+      ({ filename } = new FileSpec(data).serializable);
+    } else {
+      return null;
     }
 
-    const ext = filename.split(".").at(-1)?.toLowerCase();
-    switch (ext) {
-      case "mp4":
-      case "m4v":
-        return "video/mp4";
-      case "webm":
-        return "video/webm";
-      case "ogv":
-        return "video/ogg";
-      case "mov":
-        return "video/quicktime";
-      case "mp3":
-        return "audio/mpeg";
-      case "m4a":
-        return "audio/mp4";
-      case "wav":
-        return "audio/wav";
-      case "oga":
-      case "ogg":
-        return "audio/ogg";
-      default:
-        return null;
+    const contentType = MediaAnnotation._getContentType(
+      assetDict,
+      filename,
+      explicitType
+    );
+    if (!contentType) {
+      return null;
     }
+    return {
+      assetRef: rawData instanceof Ref ? rawData : null,
+      assetDict,
+      filename,
+      contentType,
+    };
   }
 }
 
diff --git a/src/core/file_spec.js b/src/core/file_spec.js
index 6409dc504..4dcef9476 100644
--- a/src/core/file_spec.js
+++ b/src/core/file_spec.js
@@ -107,6 +107,16 @@ class FileSpec {
     return null;
   }
 
+  /**
+   * Whether a file specification carries an embedded file we can read.
+   *
+   * @param {Dict} fileSpecDict
+   * @returns {boolean}
+   */
+  static hasEmbeddedFile(fileSpecDict) {
+    return this.pickPlatformItem(fileSpecDict.get("EF")) instanceof BaseStream;
+  }
+
   /**
    * Read attachment bytes from a file-spec dictionary.
    *
diff --git a/src/display/annotation_layer.js b/src/display/annotation_layer.js
index 991edc5d2..a7fdadef0 100644
--- a/src/display/annotation_layer.js
+++ b/src/display/annotation_layer.js
@@ -158,8 +158,11 @@ class AnnotationElementFactory {
       case AnnotationType.FILEATTACHMENT:
         return new FileAttachmentAnnotationElement(parameters);
 
+      // A Screen annotation with a rendition action plays embedded media the
+      // same way RichMedia does (see `MediaAnnotation` in the core layer).
       case AnnotationType.RICHMEDIA:
-        return new RichMediaAnnotationElement(parameters);
+      case AnnotationType.SCREEN:
+        return new MediaAnnotationElement(parameters);
 
       default:
         return new AnnotationElement(parameters);
@@ -403,7 +406,7 @@ class AnnotationElement {
     if (
       !(this instanceof WidgetAnnotationElement) &&
       !(this instanceof LinkAnnotationElement) &&
-      !(this instanceof RichMediaAnnotationElement)
+      !(this instanceof MediaAnnotationElement)
     ) {
       container.tabIndex = 0;
     }
@@ -3800,7 +3803,7 @@ class FileAttachmentAnnotationElement extends AnnotationElement {
   }
 }
 
-class RichMediaAnnotationElement extends AnnotationElement {
+class MediaAnnotationElement extends AnnotationElement {
   #abortController = new AbortController();
 
   #contentUrl = null;
@@ -3812,14 +3815,14 @@ class RichMediaAnnotationElement extends AnnotationElement {
   }
 
   render() {
-    this.container.classList.add("richMediaAnnotation");
+    this.container.classList.add("mediaAnnotation");
 
     const { filename } = this.data.richMedia;
 
     // The annotation's appearance (a poster image) is painted on the canvas;
     // overlay a play button that loads the embedded media on demand.
     const button = document.createElement("button");
-    button.className = "richMediaPlayButton";
+    button.className = "mediaPlayButton";
     button.type = "button";
     button.title = button.ariaLabel = filename;
     button.addEventListener("click", () => this.#load(button), {
@@ -3853,7 +3856,7 @@ class RichMediaAnnotationElement extends AnnotationElement {
     const isAudio = contentType.startsWith("audio/");
     const media = document.createElement(isAudio ? "audio" : "video");
     this.#media = media;
-    media.className = "richMediaContent";
+    media.className = "mediaContent";
     this._setBackgroundColor(media);
     media.src = url;
     media.title = filename;
diff --git a/src/shared/util.js b/src/shared/util.js
index 7ab5f5cd6..9757d441d 100644
--- a/src/shared/util.js
+++ b/src/shared/util.js
@@ -175,6 +175,16 @@ const AnnotationReplyType = {
   REPLY: "R",
 };
 
+// Rendition action operations from Table 214, Section 12.6.4.13 of the PDF
+// specification (ISO 32000-1).
+const AnnotationRenditionOperation = {
+  PLAY_OR_RESUME: 0,
+  STOP: 1,
+  PAUSE: 2,
+  RESUME: 3,
+  PLAY: 4,
+};
+
 const AnnotationFlag = {
   INVISIBLE: 0x01,
   HIDDEN: 0x02,
@@ -1152,6 +1162,7 @@ export {
   AnnotationFlag,
   AnnotationMode,
   AnnotationPrefix,
+  AnnotationRenditionOperation,
   AnnotationReplyType,
   AnnotationType,
   assert,
diff --git a/test/integration/annotation_spec.mjs b/test/integration/annotation_spec.mjs
index b0c65f0f8..39ce0474d 100644
--- a/test/integration/annotation_spec.mjs
+++ b/test/integration/annotation_spec.mjs
@@ -987,8 +987,8 @@ describe("RichMedia annotation", () => {
       await Promise.all(
         pages.map(async ([browserName, page]) => {
           const annotationSelector = getAnnotationSelector("4R");
-          const buttonSelector = `${annotationSelector} .richMediaPlayButton`;
-          const videoSelector = `${annotationSelector} video.richMediaContent`;
+          const buttonSelector = `${annotationSelector} .mediaPlayButton`;
+          const videoSelector = `${annotationSelector} video.mediaContent`;
 
           // Initially only the play button (over the poster) is shown.
           await page.waitForSelector(buttonSelector, { timeout: 0 });
@@ -1008,8 +1008,8 @@ describe("RichMedia annotation", () => {
       await Promise.all(
         pages.map(async ([browserName, page]) => {
           const annotationSelector = getAnnotationSelector("5R");
-          const buttonSelector = `${annotationSelector} .richMediaPlayButton`;
-          const audioSelector = `${annotationSelector} audio.richMediaContent`;
+          const buttonSelector = `${annotationSelector} .mediaPlayButton`;
+          const audioSelector = `${annotationSelector} audio.mediaContent`;
 
           // Initially only the play button is shown.
           await page.waitForSelector(buttonSelector, { timeout: 0 });
@@ -1026,3 +1026,58 @@ describe("RichMedia annotation", () => {
     });
   });
 });
+
+describe("Screen annotation (rendition)", () => {
+  describe("multimedia_annotations.pdf", () => {
+    let pages;
+
+    beforeEach(async () => {
+      pages = await loadAndWait(
+        "multimedia_annotations.pdf",
+        getAnnotationSelector("30R")
+      );
+    });
+
+    afterEach(async () => {
+      await closePages(pages);
+    });
+
+    it("must play the rendition video when clicking the play button", async () => {
+      await Promise.all(
+        pages.map(async ([browserName, page]) => {
+          const annotationSelector = getAnnotationSelector("30R");
+          const buttonSelector = `${annotationSelector} .mediaPlayButton`;
+          const videoSelector = `${annotationSelector} video.mediaContent`;
+
+          await page.waitForSelector(buttonSelector, { visible: true });
+          await page.click(buttonSelector);
+
+          await page.waitForSelector(videoSelector, { visible: true });
+          const hasSource = await page.$eval(videoSelector, el =>
+            el.src.startsWith("blob:")
+          );
+          expect(hasSource).withContext(`In ${browserName}`).toEqual(true);
+        })
+      );
+    });
+
+    it("must play the rendition audio when clicking the play button", async () => {
+      await Promise.all(
+        pages.map(async ([browserName, page]) => {
+          const annotationSelector = getAnnotationSelector("6R");
+          const buttonSelector = `${annotationSelector} .mediaPlayButton`;
+          const audioSelector = `${annotationSelector} audio.mediaContent`;
+
+          await page.waitForSelector(buttonSelector, { visible: true });
+          await page.click(buttonSelector);
+
+          await page.waitForSelector(audioSelector, { visible: true });
+          const hasSource = await page.$eval(audioSelector, el =>
+            el.src.startsWith("blob:")
+          );
+          expect(hasSource).withContext(`In ${browserName}`).toEqual(true);
+        })
+      );
+    });
+  });
+});
diff --git a/test/integration/presentation_mode_spec.mjs b/test/integration/presentation_mode_spec.mjs
index 712068b0a..add254a45 100644
--- a/test/integration/presentation_mode_spec.mjs
+++ b/test/integration/presentation_mode_spec.mjs
@@ -214,7 +214,7 @@ describe("PDFPresentationMode", () => {
       await Promise.all(
         pages.map(async ([browserName, page]) => {
           const annotationSelector = getAnnotationSelector("4R");
-          const videoSelector = `${annotationSelector} video.richMediaContent`;
+          const videoSelector = `${annotationSelector} video.mediaContent`;
 
           await page.evaluate(() => {
             window.__richMediaNavigationCalls = 0;
@@ -229,11 +229,10 @@ describe("PDFPresentationMode", () => {
           });
 
           await enterPresentationMode(page);
-          await page.waitForSelector(
-            `${annotationSelector} .richMediaPlayButton`,
-            { visible: true }
-          );
-          await page.click(`${annotationSelector} .richMediaPlayButton`);
+          await page.waitForSelector(`${annotationSelector} .mediaPlayButton`, {
+            visible: true,
+          });
+          await page.click(`${annotationSelector} .mediaPlayButton`);
           await page.waitForSelector(videoSelector, { timeout: 0 });
 
           const result = await page.$eval(videoSelector, el => ({
diff --git a/test/unit/annotation_spec.js b/test/unit/annotation_spec.js
index ef2e5ad33..2cc238826 100644
--- a/test/unit/annotation_spec.js
+++ b/test/unit/annotation_spec.js
@@ -25,6 +25,7 @@ import {
   AnnotationEditorType,
   AnnotationFieldFlag,
   AnnotationFlag,
+  AnnotationRenditionOperation,
   AnnotationType,
   bytesToString,
   DrawOPS,
@@ -145,6 +146,23 @@ describe("annotation", function () {
     partialEvaluator = null;
   });
 
+  function createAssetDict(filename, mimeSubtype = null) {
+    let streamDict = null;
+    if (mimeSubtype) {
+      streamDict = new Dict();
+      streamDict.set("Type", Name.get("EmbeddedFile"));
+      streamDict.set("Subtype", Name.get(mimeSubtype));
+    }
+    const embeddedFileDict = new Dict();
+    embeddedFileDict.set("F", new StringStream("", streamDict));
+
+    const fileSpecDict = new Dict();
+    fileSpecDict.set("Type", Name.get("Filespec"));
+    fileSpecDict.set("EF", embeddedFileDict);
+    fileSpecDict.set("UF", filename);
+    return fileSpecDict;
+  }
+
   describe("AnnotationFactory", function () {
     it("should get id for annotation", async function () {
       const annotationDict = new Dict();
@@ -4597,23 +4615,6 @@ describe("annotation", function () {
   });
 
   describe("RichMediaAnnotation", function () {
-    function createAssetDict(filename, mimeSubtype = null) {
-      let streamDict = null;
-      if (mimeSubtype) {
-        streamDict = new Dict();
-        streamDict.set("Type", Name.get("EmbeddedFile"));
-        streamDict.set("Subtype", Name.get(mimeSubtype));
-      }
-      const embeddedFileDict = new Dict();
-      embeddedFileDict.set("F", new StringStream("", streamDict));
-
-      const fileSpecDict = new Dict();
-      fileSpecDict.set("Type", Name.get("Filespec"));
-      fileSpecDict.set("EF", embeddedFileDict);
-      fileSpecDict.set("UF", filename);
-      return fileSpecDict;
-    }
-
     function createAnnotation(contentDict, refNum) {
       const dict = new Dict();
       dict.set("Type", Name.get("Annot"));
@@ -4905,6 +4906,336 @@ describe("annotation", function () {
       expect(data.noHTML).toEqual(true);
       expect(data.richMedia).toBeUndefined();
     });
+
+    it("should not create media data for an external (non-embedded) asset", async function () {
+      const assetRef = Ref.get(180, 0);
+      // A URL file specification with no `/EF`: the bytes aren't embedded, so
+      // there's nothing pdf.js can serve.
+      const assetDict = new Dict();
+      assetDict.set("Type", Name.get("Filespec"));
+      assetDict.set("FS", Name.get("URL"));
+      assetDict.set("F", "https://example.com/demo.mp4");
+
+      const instanceDict = new Dict();
+      instanceDict.set("Subtype", Name.get("Video"));
+      instanceDict.set("Asset", assetRef);
+
+      const configDict = new Dict();
+      configDict.set("Instances", [instanceDict]);
+
+      const contentDict = new Dict();
+      contentDict.set("Configurations", [configDict]);
+
+      const annotation = createAnnotation(contentDict, 181);
+      const xref = new XRefMock([
+        { ref: assetRef, data: assetDict },
+        annotation,
+      ]);
+      assetDict.assignXref(xref);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.annotationType).toEqual(AnnotationType.RICHMEDIA);
+      expect(data.noHTML).toEqual(true);
+      expect(data.richMedia).toBeUndefined();
+    });
+  });
+
+  describe("ScreenAnnotation", function () {
+    function createMediaClipDict(fileSpecRef, contentType = null) {
+      const clipDict = new Dict();
+      clipDict.set("Type", Name.get("MediaClip"));
+      clipDict.set("S", Name.get("MCD"));
+      if (contentType) {
+        clipDict.set("CT", contentType);
+      }
+      clipDict.set("D", fileSpecRef);
+      return clipDict;
+    }
+
+    function createRenditionDict(clipDict) {
+      const renditionDict = new Dict();
+      renditionDict.set("Type", Name.get("Rendition"));
+      renditionDict.set("S", Name.get("MR"));
+      renditionDict.set("C", clipDict);
+      return renditionDict;
+    }
+
+    function createRenditionAction(
+      renditionDict,
+      refNum,
+      op = AnnotationRenditionOperation.PLAY_OR_RESUME
+    ) {
+      const actionDict = new Dict();
+      actionDict.set("Type", Name.get("Action"));
+      actionDict.set("S", Name.get("Rendition"));
+      actionDict.set("OP", op);
+      actionDict.set("AN", Ref.get(refNum, 0));
+      actionDict.set("R", renditionDict);
+      return actionDict;
+    }
+
+    function createScreenAnnotation(refNum, { action = null, aa = null } = {}) {
+      const dict = new Dict();
+      dict.set("Type", Name.get("Annot"));
+      dict.set("Subtype", Name.get("Screen"));
+      if (action) {
+        dict.set("A", action);
+      }
+      if (aa) {
+        dict.set("AA", aa);
+      }
+      return { ref: Ref.get(refNum, 0), data: dict };
+    }
+
+    it("should parse the media clip from a rendition action", async function () {
+      const fileSpecRef = Ref.get(200, 0);
+      const fileSpecDict = createAssetDict("demo.mp3");
+      const clipDict = createMediaClipDict(fileSpecRef, "audio/mpeg");
+      const rendition = createRenditionDict(clipDict);
+      const annotation = createScreenAnnotation(201, {
+        action: createRenditionAction(rendition, 201),
+      });
+
+      const xref = new XRefMock([
+        { ref: fileSpecRef, data: fileSpecDict },
+        annotation,
+      ]);
+      fileSpecDict.assignXref(xref);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.annotationType).toEqual(AnnotationType.SCREEN);
+      expect(data.noHTML).toEqual(false);
+      expect(data.richMedia).toEqual({
+        fileId: "attachmentRef:200R",
+        filename: "demo.mp3",
+        contentType: "audio/mpeg",
+      });
+    });
+
+    it("should derive the content type from the file extension", async function () {
+      const fileSpecRef = Ref.get(210, 0);
+      const fileSpecDict = createAssetDict("demo.mp4");
+      // No `/CT`, so the type is inferred from the filename extension.
+      const clipDict = createMediaClipDict(fileSpecRef);
+      const rendition = createRenditionDict(clipDict);
+      const annotation = createScreenAnnotation(211, {
+        action: createRenditionAction(rendition, 211),
+      });
+
+      const xref = new XRefMock([
+        { ref: fileSpecRef, data: fileSpecDict },
+        annotation,
+      ]);
+      fileSpecDict.assignXref(xref);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.richMedia).toEqual({
+        fileId: "attachmentRef:210R",
+        filename: "demo.mp4",
+        contentType: "video/mp4",
+      });
+    });
+
+    it("should derive the content type from a direct media stream's Subtype", async function () {
+      // `/D` is the embedded media stream itself (not a file spec), and `/N` is
+      // a display label with no extension, so the type comes from the stream's
+      // own `/Subtype`.
+      const streamRef = Ref.get(250, 0);
+      const streamDict = new Dict();
+      streamDict.set("Subtype", Name.get("video/mp4"));
+      const mediaStream = new StringStream("", streamDict);
+
+      const clipDict = new Dict();
+      clipDict.set("Type", Name.get("MediaClip"));
+      clipDict.set("S", Name.get("MCD"));
+      clipDict.set("N", "Intro clip");
+      clipDict.set("D", streamRef);
+
+      const rendition = createRenditionDict(clipDict);
+      const annotation = createScreenAnnotation(251, {
+        action: createRenditionAction(rendition, 251),
+      });
+
+      const xref = new XRefMock([
+        { ref: streamRef, data: mediaStream },
+        annotation,
+      ]);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.richMedia).toEqual({
+        fileId: "attachmentRef:250R",
+        filename: "Intro clip",
+        contentType: "video/mp4",
+      });
+    });
+
+    it("should unwrap a selector rendition to the first playable media", async function () {
+      // An empty media rendition (no clip) followed by a playable one.
+      const emptyRendition = new Dict();
+      emptyRendition.set("S", Name.get("MR"));
+
+      const fileSpecRef = Ref.get(220, 0);
+      const fileSpecDict = createAssetDict("demo.mp4");
+      const rendition = createRenditionDict(
+        createMediaClipDict(fileSpecRef, "video/mp4")
+      );
+
+      const selectorDict = new Dict();
+      selectorDict.set("Type", Name.get("Rendition"));
+      selectorDict.set("S", Name.get("SR"));
+      selectorDict.set("R", [emptyRendition, rendition]);
+
+      const annotation = createScreenAnnotation(221, {
+        action: createRenditionAction(selectorDict, 221),
+      });
+
+      const xref = new XRefMock([
+        { ref: fileSpecRef, data: fileSpecDict },
+        annotation,
+      ]);
+      fileSpecDict.assignXref(xref);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.richMedia.filename).toEqual("demo.mp4");
+      expect(data.richMedia.contentType).toEqual("video/mp4");
+    });
+
+    it("should find a rendition action in the additional-actions dictionary", async function () {
+      const fileSpecRef = Ref.get(230, 0);
+      const fileSpecDict = createAssetDict("demo.mp4");
+      const rendition = createRenditionDict(
+        createMediaClipDict(fileSpecRef, "video/mp4")
+      );
+
+      // No `/A`; the rendition action is a page-visible additional action.
+      const aa = new Dict();
+      aa.set("PV", createRenditionAction(rendition, 231));
+
+      const annotation = createScreenAnnotation(231, { aa });
+
+      const xref = new XRefMock([
+        { ref: fileSpecRef, data: fileSpecDict },
+        annotation,
+      ]);
+      fileSpecDict.assignXref(xref);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.richMedia.filename).toEqual("demo.mp4");
+      expect(data.richMedia.contentType).toEqual("video/mp4");
+    });
+
+    it("should ignore a non-play rendition action", async function () {
+      const fileSpecRef = Ref.get(235, 0);
+      const fileSpecDict = createAssetDict("demo.mp4");
+      const rendition = createRenditionDict(
+        createMediaClipDict(fileSpecRef, "video/mp4")
+      );
+
+      const annotation = createScreenAnnotation(236, {
+        action: createRenditionAction(
+          rendition,
+          236,
+          AnnotationRenditionOperation.PAUSE
+        ),
+      });
+
+      const xref = new XRefMock([
+        { ref: fileSpecRef, data: fileSpecDict },
+        annotation,
+      ]);
+      fileSpecDict.assignXref(xref);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.annotationType).toEqual(AnnotationType.SCREEN);
+      expect(data.noHTML).toEqual(true);
+      expect(data.richMedia).toBeUndefined();
+    });
+
+    it("should not create media data for an external file specification", async function () {
+      const fileSpecRef = Ref.get(237, 0);
+      const fileSpecDict = new Dict();
+      fileSpecDict.set("Type", Name.get("Filespec"));
+      fileSpecDict.set("FS", Name.get("URL"));
+      fileSpecDict.set("F", "https://example.com/demo.mp4");
+
+      const rendition = createRenditionDict(
+        createMediaClipDict(fileSpecRef, "video/mp4")
+      );
+      const annotation = createScreenAnnotation(238, {
+        action: createRenditionAction(rendition, 238),
+      });
+
+      const xref = new XRefMock([
+        { ref: fileSpecRef, data: fileSpecDict },
+        annotation,
+      ]);
+      fileSpecDict.assignXref(xref);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.annotationType).toEqual(AnnotationType.SCREEN);
+      expect(data.noHTML).toEqual(true);
+      expect(data.richMedia).toBeUndefined();
+    });
+
+    it("should not create media data without a rendition action", async function () {
+      // A URI action, not a rendition; the Screen carries no playable media.
+      const uriAction = new Dict();
+      uriAction.set("S", Name.get("URI"));
+      uriAction.set("URI", "https://example.com");
+      const annotation = createScreenAnnotation(241, { action: uriAction });
+
+      const xref = new XRefMock([annotation]);
+
+      const { data } = await AnnotationFactory.create(
+        xref,
+        annotation.ref,
+        annotationGlobalsMock,
+        idFactoryMock
+      );
+      expect(data.annotationType).toEqual(AnnotationType.SCREEN);
+      expect(data.noHTML).toEqual(true);
+      expect(data.richMedia).toBeUndefined();
+    });
   });
 
   describe("PopupAnnotation", function () {
diff --git a/web/annotation_layer_builder.css b/web/annotation_layer_builder.css
index f5075ad26..af3a2af24 100644
--- a/web/annotation_layer_builder.css
+++ b/web/annotation_layer_builder.css
@@ -328,19 +328,19 @@
     width: 100%;
   }
 
-  .richMediaAnnotation {
+  .mediaAnnotation {
     pointer-events: auto;
 
-    .richMediaContent {
+    .mediaContent {
       width: 100%;
     }
 
-    video.richMediaContent {
+    video.mediaContent {
       height: 100%;
       object-fit: contain;
     }
 
-    audio.richMediaContent {
+    audio.mediaContent {
       /* An `<audio>` element has a fixed-height control bar; pin it to the
          bottom of the section so it lines up with where `<video>` shows its
          own bar. */
@@ -348,7 +348,7 @@
       bottom: 0;
     }
 
-    .richMediaPlayButton {
+    .mediaPlayButton {
       height: 100%;
       width: 100%;
       margin: 0;
diff --git a/web/annotation_layer_builder.js b/web/annotation_layer_builder.js
index d777d2992..94d98db45 100644
--- a/web/annotation_layer_builder.js
+++ b/web/annotation_layer_builder.js
@@ -285,7 +285,7 @@ class AnnotationLayerBuilder {
     for (const section of this.div.childNodes) {
       if (
         section.hasAttribute("data-internal-link") ||
-        section.classList.contains("richMediaAnnotation")
+        section.classList.contains("mediaAnnotation")
       ) {
         continue;
       }
diff --git a/web/pdf_presentation_mode.js b/web/pdf_presentation_mode.js
index 15eba52c6..44b327b72 100644
--- a/web/pdf_presentation_mode.js
+++ b/web/pdf_presentation_mode.js
@@ -126,7 +126,7 @@ class PDFPresentationMode {
     if (!this.active) {
       return;
     }
-    if (evt.target.closest?.(".richMediaAnnotation")) {
+    if (evt.target.closest?.(".mediaAnnotation")) {
       return;
     }
     evt.preventDefault();
@@ -250,7 +250,7 @@ class PDFPresentationMode {
       return;
     }
     // Allow interacting with embedded media controls rather than advancing.
-    if (evt.target.closest?.(".richMediaAnnotation")) {
+    if (evt.target.closest?.(".mediaAnnotation")) {
       return;
     }
     // Unless an internal link was clicked, advance one page.
@@ -300,7 +300,7 @@ class PDFPresentationMode {
     if (!this.active) {
       return;
     }
-    if (evt.target.closest?.(".richMediaAnnotation")) {
+    if (evt.target.closest?.(".mediaAnnotation")) {
       this.touchSwipeState = null;
       return;
     }
diff --git a/web/viewer.css b/web/viewer.css
index 373ee75a6..edb1e724b 100644
--- a/web/viewer.css
+++ b/web/viewer.css
@@ -237,7 +237,7 @@ body {
 }
 
 .pdfPresentationMode:fullscreen
-  section:not([data-internal-link], .richMediaAnnotation) {
+  section:not([data-internal-link], .mediaAnnotation) {
   pointer-events: none;
 }