@storyteller-platform/align 0.1.41 → 0.1.48

package/dist/cli/bin.cjs

@@ -359,13 +359,12 @@ async function main() {
           if (parsed.textRef === "id-fragment") {
             logger.info("Marking up EPUB...");
             startProgressBar();
-            const markedup2 = parsed.markedup ?? (0, import_node_path.join)(os.tmpdir(), `stalign-markedup-${(0, import_node_crypto.randomUUID)()}.epub`);
             if (!parsed.markedup) {
               stack.defer(() => {
-                (0, import_node_fs.rmSync)(markedup2, { recursive: true, force: true });
+                (0, import_node_fs.rmSync)(markedup, { recursive: true, force: true });
               });
             }
-            const markupTiming = await (0, import_markup.markup)(input, markedup2, {
+            const markupTiming = await (0, import_markup.markup)(input, markedup, {
               granularity: parsed.granularity,
               primaryLocale,
               logger,
@@ -376,12 +375,12 @@ async function main() {
               }
             });
             resetProgressBar();
-            logger.info(`Markup complete, marked up EPUB saved to ${markedup2}.`);
+            logger.info(`Markup complete, marked up EPUB saved to ${markedup}.`);
             if (parsed.time) {
               markupTiming.print();
             }
           } else {
-            logger.info("Skipping markup, text-range-type set to text-fragment");
+            logger.info("Skipping markup, text-ref set to text-fragment");
           }
           logger.info("Aligning EPUB with audiobook...");
           startProgressBar();

package/dist/cli/bin.js

@@ -312,13 +312,12 @@ async function main() {
           if (parsed.textRef === "id-fragment") {
             logger.info("Marking up EPUB...");
             startProgressBar();
-            const markedup2 = parsed.markedup ?? join(os.tmpdir(), `stalign-markedup-${randomUUID()}.epub`);
             if (!parsed.markedup) {
               stack.defer(() => {
-                rmSync(markedup2, { recursive: true, force: true });
+                rmSync(markedup, { recursive: true, force: true });
               });
             }
-            const markupTiming = await markup(input, markedup2, {
+            const markupTiming = await markup(input, markedup, {
               granularity: parsed.granularity,
               primaryLocale,
               logger,
@@ -329,12 +328,12 @@ async function main() {
               }
             });
             resetProgressBar();
-            logger.info(`Markup complete, marked up EPUB saved to ${markedup2}.`);
+            logger.info(`Markup complete, marked up EPUB saved to ${markedup}.`);
             if (parsed.time) {
               markupTiming.print();
             }
           } else {
-            logger.info("Skipping markup, text-range-type set to text-fragment");
+            logger.info("Skipping markup, text-ref set to text-fragment");
           }
           logger.info("Aligning EPUB with audiobook...");
           startProgressBar();

package/dist/common/ffmpeg.cjs

@@ -28,8 +28,11 @@ var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__ge
 var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
 var ffmpeg_exports = {};
 __export(ffmpeg_exports, {
+  MP3_CBR_BITRATES: () => MP3_CBR_BITRATES,
   getTrackDuration: () => getTrackDuration,
   getTrackInfo: () => getTrackInfo,
+  isVbrMp3: () => isVbrMp3,
+  selectCbrBitrate: () => selectCbrBitrate,
   splitFile: () => splitFile,
   transcodeFile: () => transcodeFile
 });
@@ -43,14 +46,11 @@ var import_mime = require("../process/mime.cjs");
 var import_shell = require("./shell.cjs");
 const execPromise = (0, import_node_util.promisify)(import_node_child_process.exec);
 async function execCmd(command, logger, signal) {
-  let stdout = "";
-  let stderr = "";
   try {
-    ;
-    ({ stdout, stderr } = await execPromise(command, {
+    const { stdout } = await execPromise(command, {
       maxBuffer: 50 * 1024 * 1024,
       signal: signal ?? void 0
-    }));
+    });
     return stdout;
   } catch (error) {
     if (error instanceof RangeError && error.message.includes("stdout maxBuffer length exceeded")) {
@@ -58,14 +58,16 @@ async function execCmd(command, logger, signal) {
         "stdout maxBuffer length exceeded. This likely means that youre trying to process a very large file, and the ffmpeg process is running out of memory. Maybe check the image size of your cover art."
       );
     }
+    const execErr = error;
     logger?.error(error);
-    logger?.info(stdout);
-    throw new Error(stderr);
+    if (execErr.stdout) logger?.info(execErr.stdout);
+    const errorDetail = execErr.stderr || execErr.stdout || `Command failed: ${command}`;
+    throw new Error(errorDetail);
   }
 }
 const getTrackInfo = (0, import_memoize.default)(async function getTrackInfo2(path, logger) {
   const stdout = await execCmd(
-    `ffprobe -i ${(0, import_shell.quotePath)(path)} -show_format -of json`,
+    `ffprobe -v error -i ${(0, import_shell.quotePath)(path)} -show_format -of json`,
     logger
   );
   const info = JSON.parse(stdout);
@@ -75,6 +77,52 @@ async function getTrackDuration(path, logger) {
   const info = await getTrackInfo(path, logger);
   return info["duration"];
 }
+const MP3_CBR_BITRATES = [
+  64e3,
+  8e4,
+  96e3,
+  112e3,
+  128e3,
+  16e4,
+  192e3,
+  224e3,
+  256e3,
+  32e4
+];
+const VBR_PROBE_PACKET_COUNT = 50;
+const MP3_CBR_MAX_DISTINCT_SIZES = 2;
+const VBR_PROBE_MIN_SEEKABLE_SECONDS = 180;
+async function probeAudioDuration(path) {
+  const stdout = await execCmd(
+    `ffprobe -i ${(0, import_shell.quotePath)(path)} -v error -show_entries format=duration -output_format json`
+  );
+  const { format } = JSON.parse(stdout);
+  const duration = Number(format?.duration);
+  return Number.isFinite(duration) && duration > 0 ? duration : null;
+}
+async function probePacketSizes(path, startSeconds) {
+  const interval = startSeconds > 0 ? `${startSeconds}%+#${VBR_PROBE_PACKET_COUNT}` : `%+#${VBR_PROBE_PACKET_COUNT}`;
+  const stdout = await execCmd(
+    `ffprobe -i ${(0, import_shell.quotePath)(path)} -v error -select_streams a:0 -read_intervals "${interval}" -show_entries packet=size -output_format json`
+  );
+  const { packets } = JSON.parse(stdout);
+  return (packets ?? []).map((packet) => Number(packet.size)).filter((size) => Number.isFinite(size) && size > 0);
+}
+async function isVbrMp3(path) {
+  if ((0, import_node_path.extname)(path).toLowerCase() !== ".mp3") return false;
+  const duration = await probeAudioDuration(path);
+  const startSeconds = duration && duration > VBR_PROBE_MIN_SEEKABLE_SECONDS ? Math.floor(duration / 3) : 0;
+  let sizes = await probePacketSizes(path, startSeconds);
+  if (sizes.length === 0 && startSeconds > 0) {
+    sizes = await probePacketSizes(path, 0);
+  }
+  if (sizes.length === 0) return false;
+  const distinctSizes = new Set(sizes).size;
+  return distinctSizes > MP3_CBR_MAX_DISTINCT_SIZES;
+}
+function selectCbrBitrate(averageBitrate) {
+  return MP3_CBR_BITRATES.find((tier) => tier >= averageBitrate) ?? MP3_CBR_BITRATES.at(-1) ?? MP3_CBR_BITRATES[0];
+}
 function parseTrackInfo(format) {
   return {
     filename: format.filename,
@@ -137,15 +185,16 @@ async function constructExtractCoverArtCommand(source, destExtension) {
   ];
   return `${command} ${args.join(" ")} | `;
 }
-function commonFfmpegArguments(sourceExtension, destExtension, codec, bitrate) {
+function commonFfmpegArguments(options) {
+  const { sourceExtension, destExtension, codec, bitrate } = options;
   const args = ["-vn"];
   if (codec) {
-    args.push(
-      "-c:a",
-      codec,
-      ...codec === "libopus" ? ["-b:a", bitrate && /^\d+[kK]$/i.test(bitrate) ? bitrate : "32K"] : [],
-      ...codec === "libmp3lame" && bitrate ? ["-q:a", bitrate] : []
-    );
+    args.push("-c:a", codec);
+    if (codec === "libopus") {
+      args.push("-b:a", bitrate && /^\d+[kK]$/i.test(bitrate) ? bitrate : "32K");
+    } else if (codec === "libmp3lame" && bitrate) {
+      args.push("-b:a", bitrate);
+    }
   } else if ((0, import_mime.areSameType)(sourceExtension, destExtension) || destExtension == ".mp4") {
     args.push("-c:a", "copy");
   }
@@ -169,12 +218,12 @@ async function splitFile(input, output, start, end, encoding, signal, logger) {
     end,
     "-i",
     (0, import_shell.quotePath)(input),
-    ...commonFfmpegArguments(
-      (0, import_node_path.extname)(input),
-      (0, import_node_path.extname)(output),
-      encoding?.codec ?? null,
-      encoding?.bitrate ?? null
-    ),
+    ...commonFfmpegArguments({
+      sourceExtension: (0, import_node_path.extname)(input),
+      destExtension: (0, import_node_path.extname)(output),
+      codec: encoding?.codec ?? null,
+      bitrate: encoding?.bitrate ?? null
+    }),
     (0, import_shell.quotePath)(output)
   ];
   const coverArtCommand = await constructExtractCoverArtCommand(
@@ -204,12 +253,12 @@ async function transcodeFile(input, output, encoding, signal, logger) {
     "-nostdin",
     "-i",
     (0, import_shell.quotePath)(input),
-    ...commonFfmpegArguments(
-      (0, import_node_path.extname)(input),
-      (0, import_node_path.extname)(output),
-      encoding?.codec ?? null,
-      encoding?.bitrate ?? null
-    ),
+    ...commonFfmpegArguments({
+      sourceExtension: (0, import_node_path.extname)(input),
+      destExtension: (0, import_node_path.extname)(output),
+      codec: encoding?.codec ?? null,
+      bitrate: encoding?.bitrate ?? null
+    }),
     (0, import_shell.quotePath)(output)
   ];
   const coverArtCommand = await constructExtractCoverArtCommand(
@@ -225,8 +274,11 @@ async function transcodeFile(input, output, encoding, signal, logger) {
 }
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
+  MP3_CBR_BITRATES,
   getTrackDuration,
   getTrackInfo,
+  isVbrMp3,
+  selectCbrBitrate,
   splitFile,
   transcodeFile
 });

package/dist/common/ffmpeg.d.cts

@@ -3,6 +3,22 @@ import { AudioEncoding } from '../process/AudioEncoding.cjs';
 
 declare const getTrackInfo: (path: string, logger?: Logger) => Promise<TrackInfo>;
 declare function getTrackDuration(path: string, logger?: Logger): Promise<number>;
+/**
+ * CBR bitrates (bps) offered for MP3 output, roughly matching LAME -V9..-V0
+ */
+declare const MP3_CBR_BITRATES: readonly [64000, 80000, 96000, 112000, 128000, 160000, 192000, 224000, 256000, 320000];
+/**
+ * Detect whether an MP3 file uses a variable bitrate
+ * Does this by sampling the first few packets and checking if the sizes are different
+ * CBR MP3 files will have the same packet size for the entire file
+ *
+ * Can't really trust the reported bitrate to tell CBR from VBR
+ * LAME writes a Xing header carrying the *average* bitrate,
+ * which ffprobe surfaces as a normal per-stream `bit_rate`,
+ * so a VBR file looks identical to a CBR one by that measure.
+ */
+declare function isVbrMp3(path: string): Promise<boolean>;
+declare function selectCbrBitrate(averageBitrate: number): number;
 type TrackInfo = {
     filename: string;
     nbStreams: number;
@@ -30,4 +46,4 @@ type TrackInfo = {
 declare function splitFile(input: string, output: string, start: number, end: number, encoding?: AudioEncoding | null, signal?: AbortSignal | null, logger?: Logger | null): Promise<boolean>;
 declare function transcodeFile(input: string, output: string, encoding?: AudioEncoding | null, signal?: AbortSignal | null, logger?: Logger | null): Promise<true | undefined>;
 
-export { getTrackDuration, getTrackInfo, splitFile, transcodeFile };
+export { MP3_CBR_BITRATES, getTrackDuration, getTrackInfo, isVbrMp3, selectCbrBitrate, splitFile, transcodeFile };

package/dist/common/ffmpeg.d.ts

@@ -3,6 +3,22 @@ import { AudioEncoding } from '../process/AudioEncoding.js';
 
 declare const getTrackInfo: (path: string, logger?: Logger) => Promise<TrackInfo>;
 declare function getTrackDuration(path: string, logger?: Logger): Promise<number>;
+/**
+ * CBR bitrates (bps) offered for MP3 output, roughly matching LAME -V9..-V0
+ */
+declare const MP3_CBR_BITRATES: readonly [64000, 80000, 96000, 112000, 128000, 160000, 192000, 224000, 256000, 320000];
+/**
+ * Detect whether an MP3 file uses a variable bitrate
+ * Does this by sampling the first few packets and checking if the sizes are different
+ * CBR MP3 files will have the same packet size for the entire file
+ *
+ * Can't really trust the reported bitrate to tell CBR from VBR
+ * LAME writes a Xing header carrying the *average* bitrate,
+ * which ffprobe surfaces as a normal per-stream `bit_rate`,
+ * so a VBR file looks identical to a CBR one by that measure.
+ */
+declare function isVbrMp3(path: string): Promise<boolean>;
+declare function selectCbrBitrate(averageBitrate: number): number;
 type TrackInfo = {
     filename: string;
     nbStreams: number;
@@ -30,4 +46,4 @@ type TrackInfo = {
 declare function splitFile(input: string, output: string, start: number, end: number, encoding?: AudioEncoding | null, signal?: AbortSignal | null, logger?: Logger | null): Promise<boolean>;
 declare function transcodeFile(input: string, output: string, encoding?: AudioEncoding | null, signal?: AbortSignal | null, logger?: Logger | null): Promise<true | undefined>;
 
-export { getTrackDuration, getTrackInfo, splitFile, transcodeFile };
+export { MP3_CBR_BITRATES, getTrackDuration, getTrackInfo, isVbrMp3, selectCbrBitrate, splitFile, transcodeFile };

package/dist/common/ffmpeg.js

@@ -8,14 +8,11 @@ import { areSameType } from "../process/mime.js";
 import { quotePath } from "./shell.js";
 const execPromise = promisify(exec);
 async function execCmd(command, logger, signal) {
-  let stdout = "";
-  let stderr = "";
   try {
-    ;
-    ({ stdout, stderr } = await execPromise(command, {
+    const { stdout } = await execPromise(command, {
       maxBuffer: 50 * 1024 * 1024,
       signal: signal ?? void 0
-    }));
+    });
     return stdout;
   } catch (error) {
     if (error instanceof RangeError && error.message.includes("stdout maxBuffer length exceeded")) {
@@ -23,14 +20,16 @@ async function execCmd(command, logger, signal) {
         "stdout maxBuffer length exceeded. This likely means that youre trying to process a very large file, and the ffmpeg process is running out of memory. Maybe check the image size of your cover art."
       );
     }
+    const execErr = error;
     logger?.error(error);
-    logger?.info(stdout);
-    throw new Error(stderr);
+    if (execErr.stdout) logger?.info(execErr.stdout);
+    const errorDetail = execErr.stderr || execErr.stdout || `Command failed: ${command}`;
+    throw new Error(errorDetail);
   }
 }
 const getTrackInfo = memoize(async function getTrackInfo2(path, logger) {
   const stdout = await execCmd(
-    `ffprobe -i ${quotePath(path)} -show_format -of json`,
+    `ffprobe -v error -i ${quotePath(path)} -show_format -of json`,
     logger
   );
   const info = JSON.parse(stdout);
@@ -40,6 +39,52 @@ async function getTrackDuration(path, logger) {
   const info = await getTrackInfo(path, logger);
   return info["duration"];
 }
+const MP3_CBR_BITRATES = [
+  64e3,
+  8e4,
+  96e3,
+  112e3,
+  128e3,
+  16e4,
+  192e3,
+  224e3,
+  256e3,
+  32e4
+];
+const VBR_PROBE_PACKET_COUNT = 50;
+const MP3_CBR_MAX_DISTINCT_SIZES = 2;
+const VBR_PROBE_MIN_SEEKABLE_SECONDS = 180;
+async function probeAudioDuration(path) {
+  const stdout = await execCmd(
+    `ffprobe -i ${quotePath(path)} -v error -show_entries format=duration -output_format json`
+  );
+  const { format } = JSON.parse(stdout);
+  const duration = Number(format?.duration);
+  return Number.isFinite(duration) && duration > 0 ? duration : null;
+}
+async function probePacketSizes(path, startSeconds) {
+  const interval = startSeconds > 0 ? `${startSeconds}%+#${VBR_PROBE_PACKET_COUNT}` : `%+#${VBR_PROBE_PACKET_COUNT}`;
+  const stdout = await execCmd(
+    `ffprobe -i ${quotePath(path)} -v error -select_streams a:0 -read_intervals "${interval}" -show_entries packet=size -output_format json`
+  );
+  const { packets } = JSON.parse(stdout);
+  return (packets ?? []).map((packet) => Number(packet.size)).filter((size) => Number.isFinite(size) && size > 0);
+}
+async function isVbrMp3(path) {
+  if (extname(path).toLowerCase() !== ".mp3") return false;
+  const duration = await probeAudioDuration(path);
+  const startSeconds = duration && duration > VBR_PROBE_MIN_SEEKABLE_SECONDS ? Math.floor(duration / 3) : 0;
+  let sizes = await probePacketSizes(path, startSeconds);
+  if (sizes.length === 0 && startSeconds > 0) {
+    sizes = await probePacketSizes(path, 0);
+  }
+  if (sizes.length === 0) return false;
+  const distinctSizes = new Set(sizes).size;
+  return distinctSizes > MP3_CBR_MAX_DISTINCT_SIZES;
+}
+function selectCbrBitrate(averageBitrate) {
+  return MP3_CBR_BITRATES.find((tier) => tier >= averageBitrate) ?? MP3_CBR_BITRATES.at(-1) ?? MP3_CBR_BITRATES[0];
+}
 function parseTrackInfo(format) {
   return {
     filename: format.filename,
@@ -102,15 +147,16 @@ async function constructExtractCoverArtCommand(source, destExtension) {
   ];
   return `${command} ${args.join(" ")} | `;
 }
-function commonFfmpegArguments(sourceExtension, destExtension, codec, bitrate) {
+function commonFfmpegArguments(options) {
+  const { sourceExtension, destExtension, codec, bitrate } = options;
   const args = ["-vn"];
   if (codec) {
-    args.push(
-      "-c:a",
-      codec,
-      ...codec === "libopus" ? ["-b:a", bitrate && /^\d+[kK]$/i.test(bitrate) ? bitrate : "32K"] : [],
-      ...codec === "libmp3lame" && bitrate ? ["-q:a", bitrate] : []
-    );
+    args.push("-c:a", codec);
+    if (codec === "libopus") {
+      args.push("-b:a", bitrate && /^\d+[kK]$/i.test(bitrate) ? bitrate : "32K");
+    } else if (codec === "libmp3lame" && bitrate) {
+      args.push("-b:a", bitrate);
+    }
   } else if (areSameType(sourceExtension, destExtension) || destExtension == ".mp4") {
     args.push("-c:a", "copy");
   }
@@ -134,12 +180,12 @@ async function splitFile(input, output, start, end, encoding, signal, logger) {
     end,
     "-i",
     quotePath(input),
-    ...commonFfmpegArguments(
-      extname(input),
-      extname(output),
-      encoding?.codec ?? null,
-      encoding?.bitrate ?? null
-    ),
+    ...commonFfmpegArguments({
+      sourceExtension: extname(input),
+      destExtension: extname(output),
+      codec: encoding?.codec ?? null,
+      bitrate: encoding?.bitrate ?? null
+    }),
     quotePath(output)
   ];
   const coverArtCommand = await constructExtractCoverArtCommand(
@@ -169,12 +215,12 @@ async function transcodeFile(input, output, encoding, signal, logger) {
     "-nostdin",
     "-i",
     quotePath(input),
-    ...commonFfmpegArguments(
-      extname(input),
-      extname(output),
-      encoding?.codec ?? null,
-      encoding?.bitrate ?? null
-    ),
+    ...commonFfmpegArguments({
+      sourceExtension: extname(input),
+      destExtension: extname(output),
+      codec: encoding?.codec ?? null,
+      bitrate: encoding?.bitrate ?? null
+    }),
     quotePath(output)
   ];
   const coverArtCommand = await constructExtractCoverArtCommand(
@@ -189,8 +235,11 @@ async function transcodeFile(input, output, encoding, signal, logger) {
   return true;
 }
 export {
+  MP3_CBR_BITRATES,
   getTrackDuration,
   getTrackInfo,
+  isVbrMp3,
+  selectCbrBitrate,
   splitFile,
   transcodeFile
 };

package/dist/process/processAudiobook.cjs

@@ -126,6 +126,27 @@ async function processAudiobook(input, output, options) {
   );
   return timing;
 }
+async function resolveVbrEncoding(filepath, userEncoding, logger) {
+  if (userEncoding?.codec && userEncoding.codec !== "libmp3lame") {
+    return userEncoding;
+  }
+  const sourceIsMp3 = (0, import_node_path.extname)(filepath).toLowerCase() === ".mp3";
+  if (!userEncoding?.codec && !sourceIsMp3) {
+    return userEncoding;
+  }
+  if (!userEncoding?.codec && !await (0, import_ffmpeg.isVbrMp3)(filepath)) {
+    return userEncoding;
+  }
+  const trackInfo = await (0, import_ffmpeg.getTrackInfo)(filepath, logger ?? void 0);
+  const targetBitrate = (0, import_ffmpeg.selectCbrBitrate)(trackInfo.bitRate);
+  logger?.info(
+    `Forcing CBR MP3 for ${filepath} (avg ${trackInfo.bitRate}bps) at ${targetBitrate / 1e3}k`
+  );
+  return {
+    codec: "libmp3lame",
+    bitrate: `${targetBitrate / 1e3}k`
+  };
+}
 async function processFile(input, output, prefix, options) {
   var _stack = [];
   try {
@@ -144,13 +165,26 @@ async function processFile(input, output, prefix, options) {
       options.signal,
       options.logger
     );
+    const vbrEncodings = /* @__PURE__ */ new Map();
+    const uniqueFilepaths = [...new Set(ranges.map((r) => r.filepath))];
+    await Promise.all(
+      uniqueFilepaths.map(async (filepath) => {
+        const result = await resolveVbrEncoding(
+          filepath,
+          options.encoding,
+          options.logger
+        );
+        vbrEncodings.set(filepath, result);
+      })
+    );
     await Promise.all(
       ranges.map(async (range, index) => {
         var _stack2 = [];
         try {
+          const effectiveEncoding = vbrEncodings.has(range.filepath) ? vbrEncodings.get(range.filepath) : options.encoding;
           const outputExtension = determineExtension(
             range.filepath,
-            options.encoding?.codec
+            effectiveEncoding?.codec
           );
           const outputFilename = `${prefix}${(index + 1).toString().padStart(5, "0")}${outputExtension}`;
           const outputFilepath = (0, import_node_path.join)(output, outputFilename);
@@ -168,7 +202,7 @@ async function processFile(input, output, prefix, options) {
                 outputFilepath,
                 range.start,
                 range.end,
-                options.encoding,
+                effectiveEncoding,
                 options.signal,
                 options.logger
               );

package/dist/process/processAudiobook.js

@@ -19,7 +19,12 @@ import {
   createAggregator,
   createTiming
 } from "@storyteller-platform/ghost-story";
-import { splitFile } from "../common/ffmpeg.js";
+import {
+  getTrackInfo,
+  isVbrMp3,
+  selectCbrBitrate,
+  splitFile
+} from "../common/ffmpeg.js";
 import { getSafeChapterRanges } from "./ranges.js";
 async function processAudiobook(input, output, options) {
   const timing = createAggregator();
@@ -73,6 +78,27 @@ async function processAudiobook(input, output, options) {
   );
   return timing;
 }
+async function resolveVbrEncoding(filepath, userEncoding, logger) {
+  if (userEncoding?.codec && userEncoding.codec !== "libmp3lame") {
+    return userEncoding;
+  }
+  const sourceIsMp3 = extname(filepath).toLowerCase() === ".mp3";
+  if (!userEncoding?.codec && !sourceIsMp3) {
+    return userEncoding;
+  }
+  if (!userEncoding?.codec && !await isVbrMp3(filepath)) {
+    return userEncoding;
+  }
+  const trackInfo = await getTrackInfo(filepath, logger ?? void 0);
+  const targetBitrate = selectCbrBitrate(trackInfo.bitRate);
+  logger?.info(
+    `Forcing CBR MP3 for ${filepath} (avg ${trackInfo.bitRate}bps) at ${targetBitrate / 1e3}k`
+  );
+  return {
+    codec: "libmp3lame",
+    bitrate: `${targetBitrate / 1e3}k`
+  };
+}
 async function processFile(input, output, prefix, options) {
   var _stack = [];
   try {
@@ -91,13 +117,26 @@ async function processFile(input, output, prefix, options) {
       options.signal,
       options.logger
     );
+    const vbrEncodings = /* @__PURE__ */ new Map();
+    const uniqueFilepaths = [...new Set(ranges.map((r) => r.filepath))];
+    await Promise.all(
+      uniqueFilepaths.map(async (filepath) => {
+        const result = await resolveVbrEncoding(
+          filepath,
+          options.encoding,
+          options.logger
+        );
+        vbrEncodings.set(filepath, result);
+      })
+    );
     await Promise.all(
       ranges.map(async (range, index) => {
         var _stack2 = [];
         try {
+          const effectiveEncoding = vbrEncodings.has(range.filepath) ? vbrEncodings.get(range.filepath) : options.encoding;
           const outputExtension = determineExtension(
             range.filepath,
-            options.encoding?.codec
+            effectiveEncoding?.codec
           );
           const outputFilename = `${prefix}${(index + 1).toString().padStart(5, "0")}${outputExtension}`;
           const outputFilepath = join(output, outputFilename);
@@ -115,7 +154,7 @@ async function processFile(input, output, prefix, options) {
                 outputFilepath,
                 range.start,
                 range.end,
-                options.encoding,
+                effectiveEncoding,
                 options.signal,
                 options.logger
               );

package/dist/readium/manifest.cjs

@@ -88,10 +88,19 @@ async function generateReadiumManifest(epub, options = {}) {
   const dir = await epub.getBaseDirection();
   const epubMetadata = await epub.getMetadata();
   const vocab = await epub.getPackageVocabularyPrefixes();
-  const duration = epubMetadata.find(
-    ({ properties }) => properties["property"] === "media:duration"
-  )?.value;
-  const durationMs = duration !== void 0 ? (0, import_smil_clockvalue.default)(duration) : void 0;
+  let duration = void 0;
+  const refinesDurationMap = /* @__PURE__ */ new Map();
+  for (const dur of epubMetadata) {
+    if (dur.properties["property"] !== "media:duration") continue;
+    if (!dur.properties["refines"]) {
+      duration = dur.value ? (0, import_smil_clockvalue.default)(dur.value) / 1e3 : void 0;
+      continue;
+    }
+    const value = dur.value ? (0, import_smil_clockvalue.default)(dur.value) / 1e3 : void 0;
+    if (value) {
+      refinesDurationMap.set(dur.properties["refines"], value);
+    }
+  }
   const otherMetadata = epubMetadata.filter(
     (meta) => (meta.properties["property"]?.split(":")[0] ?? "") in vocab
   ).map((meta) => {
@@ -130,8 +139,8 @@ async function generateReadiumManifest(epub, options = {}) {
     ...dir !== "auto" && {
       readingProgression: dir === "ltr" ? import_shared.ReadingProgression.ltr : import_shared.ReadingProgression.rtl
     },
-    // TODO: is this meant to be in milliseconds (as here) or seconds?
-    ...durationMs !== void 0 && { duration: durationMs },
+    // it's seconds
+    ...duration !== void 0 && { duration },
     ...numberOfPages !== void 0 && { numberOfPages },
     otherMetadata: Object.fromEntries(otherMetadata)
   });
@@ -184,16 +193,13 @@ async function generateReadiumManifest(epub, options = {}) {
       if (!item.mediaOverlay) return link;
       const mediaOverlayItem = epubManifest[item.id];
       if (!mediaOverlayItem) return link;
-      const refinedBy = epubMetadata.find(
-        ({ properties }) => properties["property"] === "media:duration" && properties["refines"] === `#${mediaOverlayItem.id}`
-      );
-      if (!refinedBy?.value) return link;
-      const itemDuration = (0, import_smil_clockvalue.default)(refinedBy.value);
+      const duration2 = refinesDurationMap.get(`#${item.mediaOverlay}`) || refinesDurationMap.get(`#${mediaOverlayItem.id}`);
+      if (!duration2) return link;
       return new import_shared.Link({
         href: link.href,
         type: link.mediaType.string,
         ...link.properties && { properties: link.properties },
-        duration: itemDuration
+        duration: duration2
       });
     })
   );

package/dist/readium/manifest.js

@@ -71,10 +71,19 @@ async function generateReadiumManifest(epub, options = {}) {
   const dir = await epub.getBaseDirection();
   const epubMetadata = await epub.getMetadata();
   const vocab = await epub.getPackageVocabularyPrefixes();
-  const duration = epubMetadata.find(
-    ({ properties }) => properties["property"] === "media:duration"
-  )?.value;
-  const durationMs = duration !== void 0 ? clockvalue(duration) : void 0;
+  let duration = void 0;
+  const refinesDurationMap = /* @__PURE__ */ new Map();
+  for (const dur of epubMetadata) {
+    if (dur.properties["property"] !== "media:duration") continue;
+    if (!dur.properties["refines"]) {
+      duration = dur.value ? clockvalue(dur.value) / 1e3 : void 0;
+      continue;
+    }
+    const value = dur.value ? clockvalue(dur.value) / 1e3 : void 0;
+    if (value) {
+      refinesDurationMap.set(dur.properties["refines"], value);
+    }
+  }
   const otherMetadata = epubMetadata.filter(
     (meta) => (meta.properties["property"]?.split(":")[0] ?? "") in vocab
   ).map((meta) => {
@@ -113,8 +122,8 @@ async function generateReadiumManifest(epub, options = {}) {
     ...dir !== "auto" && {
       readingProgression: dir === "ltr" ? ReadingProgression.ltr : ReadingProgression.rtl
     },
-    // TODO: is this meant to be in milliseconds (as here) or seconds?
-    ...durationMs !== void 0 && { duration: durationMs },
+    // it's seconds
+    ...duration !== void 0 && { duration },
     ...numberOfPages !== void 0 && { numberOfPages },
     otherMetadata: Object.fromEntries(otherMetadata)
   });
@@ -167,16 +176,13 @@ async function generateReadiumManifest(epub, options = {}) {
       if (!item.mediaOverlay) return link;
       const mediaOverlayItem = epubManifest[item.id];
       if (!mediaOverlayItem) return link;
-      const refinedBy = epubMetadata.find(
-        ({ properties }) => properties["property"] === "media:duration" && properties["refines"] === `#${mediaOverlayItem.id}`
-      );
-      if (!refinedBy?.value) return link;
-      const itemDuration = clockvalue(refinedBy.value);
+      const duration2 = refinesDurationMap.get(`#${item.mediaOverlay}`) || refinesDurationMap.get(`#${mediaOverlayItem.id}`);
+      if (!duration2) return link;
       return new Link({
         href: link.href,
         type: link.mediaType.string,
         ...link.properties && { properties: link.properties },
-        duration: itemDuration
+        duration: duration2
       });
     })
   );

package/dist/transcribe/transcribe.cjs

@@ -101,8 +101,9 @@ async function transcribe(input, output, locale, options) {
   const engine = options.engine ?? "whisper.cpp";
   const model = options.model ?? "tiny.en";
   if (engine === "whisper.cpp") {
+    const resolvedModel = getWhisperCppModelId(locale.language, model);
     await (0, import_ghost_story.ensureWhisperInstalled)({
-      model,
+      model: resolvedModel,
       printOutput: ["debug", "info"].includes(
         options.logger?.level ?? "silent"
       ),
@@ -205,8 +206,12 @@ async function transcribeFile(input, locale, options) {
       const fallbackVariant = getCpuOverrideVariant(
         options.whisperCpuOverride ?? null
       );
+      const resolvedModel = getWhisperCppModelId(
+        sharedOptions.language,
+        options.model
+      );
       const whisperOptions = await (0, import_ghost_story.ensureWhisperInstalled)({
-        model: options.model,
+        model: resolvedModel,
         variant: fallbackVariant,
         printOutput: ["debug", "info"].includes(
           options.logger?.level ?? "silent"
@@ -218,7 +223,7 @@ async function transcribeFile(input, locale, options) {
         engine: options.engine,
         options: {
           flashAttention: true,
-          model: getWhisperCppModelId(sharedOptions.language, options.model),
+          model: resolvedModel,
           processors: options.processors,
           threads: options.threads,
           onProgress: (progress) => {

package/dist/transcribe/transcribe.js

@@ -32,8 +32,9 @@ async function transcribe(input, output, locale, options) {
   const engine = options.engine ?? "whisper.cpp";
   const model = options.model ?? "tiny.en";
   if (engine === "whisper.cpp") {
+    const resolvedModel = getWhisperCppModelId(locale.language, model);
     await ensureWhisperInstalled({
-      model,
+      model: resolvedModel,
       printOutput: ["debug", "info"].includes(
         options.logger?.level ?? "silent"
       ),
@@ -136,8 +137,12 @@ async function transcribeFile(input, locale, options) {
       const fallbackVariant = getCpuOverrideVariant(
         options.whisperCpuOverride ?? null
       );
+      const resolvedModel = getWhisperCppModelId(
+        sharedOptions.language,
+        options.model
+      );
       const whisperOptions = await ensureWhisperInstalled({
-        model: options.model,
+        model: resolvedModel,
         variant: fallbackVariant,
         printOutput: ["debug", "info"].includes(
           options.logger?.level ?? "silent"
@@ -149,7 +154,7 @@ async function transcribeFile(input, locale, options) {
         engine: options.engine,
         options: {
           flashAttention: true,
-          model: getWhisperCppModelId(sharedOptions.language, options.model),
+          model: resolvedModel,
           processors: options.processors,
           threads: options.threads,
           onProgress: (progress) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@storyteller-platform/align",
-  "version": "0.1.41",
+  "version": "0.1.48",
   "description": "A library and CLI for automatically aligning audiobooks and EPUBs to produce Media Overlays",
   "author": "Shane Friedman",
   "license": "MIT",
@@ -70,8 +70,8 @@
     "@optique/core": "^0.10.7",
     "@optique/run": "^0.10.7",
     "@readium/shared": "^2.2.0",
-    "@storyteller-platform/audiobook": "^0.4.0",
-    "@storyteller-platform/epub": "^0.6.0",
+    "@storyteller-platform/audiobook": "^0.4.1",
+    "@storyteller-platform/epub": "^0.6.2",
     "@storyteller-platform/ghost-story": "^0.1.11",
     "@storyteller-platform/transliteration": "^3.1.2",
     "chalk": "^5.4.1",