@lalalic/markcut 1.2.0 → 2.0.0

package/src/player/pipeline.mjs

@@ -189,19 +189,13 @@ function wrapWithEffects(node2, result, parentKind) {
   if (!rawEffects || rawEffects.length === 0) return result;
   const effects = rawEffects.map(normalizeEffectSpec);
   const innerStream = result.stream;
-  const innerActions = innerStream.actions ?? [];
-  const firstAction = innerActions[0] ?? {};
-  const absStart = firstAction.start ?? 0;
-  const absEnd = firstAction.end ?? result.duration;
+  const absStart = innerStream.start ?? 0;
+  const absEnd = innerStream.end ?? result.duration;
   const duration = absEnd - absStart;
   const resetStream = {
     ...innerStream,
-    actions: [{
-      ...firstAction,
-      id: uid(),
-      start: 0,
-      end: duration
-    }],
+    start: 0,
+    end: duration,
     durationInSeconds: duration
   };
   let currentStream = resetStream;
@@ -220,11 +214,8 @@ function wrapWithEffects(node2, result, parentKind) {
       animationIterationCount: spec.animationIterationCount ?? 1,
       customKeyframes: spec.customKeyframes,
       children: [currentStream],
-      actions: [{
-        id: uid(),
-        start: effStart,
-        end: effEnd
-      }],
+      start: effStart,
+      end: effEnd,
       visible: true,
       ...pickOn(node2)
     };
@@ -241,17 +232,12 @@ function compileLeaf(node2, ctx, parentKind) {
     style: node2.style,
     visible: node2.visible ?? true,
     isBackground: node2.isBackground,
-    durationInSeconds: end,
-    ...pickOn(node2)
-  };
-  const action = {
-    id: uid(),
     start,
     end,
     startFrom: node2.type === "video" || node2.type === "audio" ? node2.startFrom : void 0,
     endAt: node2.type === "video" || node2.type === "audio" ? node2.endAt : void 0,
-    loop: node2.type === "audio" ? node2.loop : void 0,
-    volume: node2.type === "video" || node2.type === "audio" || node2.type === "rhythm" ? node2.volume : void 0
+    durationInSeconds: end,
+    ...pickOn(node2)
   };
   switch (node2.type) {
     case "video": {
@@ -262,8 +248,7 @@ function compileLeaf(node2, ctx, parentKind) {
         volume: node2.volume ?? 1,
         playbackRate: node2.playbackRate,
         width: node2.width ?? 1080,
-        height: node2.height ?? 1920,
-        actions: [action]
+        height: node2.height ?? 1920
       };
       return { stream, duration: end };
     }
@@ -274,7 +259,8 @@ function compileLeaf(node2, ctx, parentKind) {
         src: node2.src,
         volume: node2.volume ?? 1,
         foreground: node2.foreground,
-        actions: [action]
+        loop: node2.loop,
+        speaker: node2.speaker
       };
       return { stream, duration: end };
     }
@@ -283,8 +269,7 @@ function compileLeaf(node2, ctx, parentKind) {
         ...base,
         type: "image",
         src: node2.src,
-        fit: node2.fit ?? "contain",
-        actions: [action]
+        fit: node2.fit ?? "contain"
       };
       return { stream, duration: end };
     }
@@ -312,8 +297,7 @@ function compileLeaf(node2, ctx, parentKind) {
         ...base,
         type: "component",
         jsx: node2.jsx,
-        data: Object.keys(bindings).length ? bindings : void 0,
-        actions: [action]
+        data: Object.keys(bindings).length ? bindings : void 0
       };
       return { stream, duration: end };
     }
@@ -324,8 +308,7 @@ function compileLeaf(node2, ctx, parentKind) {
         src: node2.src,
         volume: node2.volume ?? 1,
         spots: node2.spots,
-        children: [],
-        actions: [action]
+        children: []
       };
       return { stream, duration: end };
     }
@@ -341,7 +324,7 @@ function compileLeaf(node2, ctx, parentKind) {
         mapType: node2.mapType ?? "roadmap",
         travelMode: node2.travelMode ?? "DRIVING",
         routeMarker: node2.routeMarker ?? "\u{1F697}",
-        actions: [action]
+        googleMapsApiKey: ctx.googleMapsApiKey
       };
       return { stream, duration: end };
     }
@@ -397,6 +380,7 @@ function compileScene(node2, ctx, parentKind) {
       type: "folder",
       visible: true,
       isSeries: true,
+      start: 0,
       transition: sceneKind === "transitionSeries" ? resolved.name : void 0,
       transitionTime: sceneKind === "transitionSeries" ? resolved.time : 0.5,
       children: compiledChildren.map((c) => c.stream),
@@ -416,6 +400,7 @@ function compileScene(node2, ctx, parentKind) {
     style: node2.style,
     visible: node2.visible ?? true,
     isBackground: node2.isBackground,
+    start,
     children: sceneChildren,
     durationInSeconds: end,
     ...pickOn(node2)
@@ -447,13 +432,8 @@ function compileInclude(node2, ctx, parentKind) {
     src: node2.src,
     volume: node2.volume ?? 1,
     children: compiledChildren.map((c) => c.stream),
-    actions: [
-      {
-        id: uid(),
-        start,
-        end
-      }
-    ],
+    start,
+    end,
     durationInSeconds: end,
     ...pickOn(node2)
   };
@@ -489,14 +469,8 @@ function compileRhythm(node2, ctx, parentKind) {
     volume: node2.volume ?? 1,
     spots: node2.spots,
     children: compiledChildren,
-    actions: [
-      {
-        id: uid(),
-        start,
-        end,
-        volume: node2.volume
-      }
-    ],
+    start,
+    end,
     durationInSeconds: end,
     ...pickOn(node2)
   };
@@ -514,6 +488,7 @@ function compileContainer(node2, ctx, parentKind) {
     style: node2.style,
     visible: node2.visible ?? true,
     isBackground: node2.isBackground,
+    start: 0,
     isSeries: node2.type !== "parallel",
     transition: node2.type === "transitionSeries" ? resolved.name : void 0,
     transitionTime: node2.type === "transitionSeries" ? resolved.time : 0.5,
@@ -733,17 +708,20 @@ function resolveVariantOverrides(root, variantChain) {
   return clone;
 }
 function compileDescriptiveRoot(input, options = {}) {
+  const root = typeof input === "string" ? JSON.parse(input) : input;
+  const resolved = resolveTransition(root.layout ? root.transition ?? root.layout : root.transition, root.transitionTime);
+  const rootKind = root.layout === "parallel" ? "parallel" : "series";
+  const googleMapsApiKey = options.googleMapsApiKey ?? "";
   const ctx = {
     defaults: {
       ...DEFAULTS,
       ...options.defaults ?? {}
-    }
+    },
+    googleMapsApiKey
   };
   const registry = resolveComponentSources(input);
   warnUnregisteredComponents(input, registry);
   ensureUniqueIds(input.children, "root");
-  const rootKind = input.layout ?? "series";
-  const resolved = resolveTransition(input.transition, input.transitionTime);
   const children = compileChildren(input.children, ctx, rootKind);
   const duration = aggregateDuration(children, rootKind, resolved.time);
   const compiled = {
@@ -992,10 +970,22 @@ import { join, dirname as dirname2, resolve as resolvePath, relative } from "nod
 import { execSync } from "node:child_process";
 import { existsSync, mkdirSync } from "node:fs";
 import { dirname } from "node:path";
-var DEFAULT_TTS_CLI = process.env.MARKCUT_TTS_CLI || 'uvx edge-tts --voice "en-US-GuyNeural" --text "{input}" --write-media "{output}"';
+
+// src/config.mjs
+var MAX_IMAGE_DIMENSION = Number(process.env.MARKCUT_MAX_IMAGE_DIMENSION) || 384;
+var MAX_VIDEO_DURATION = Number(process.env.MARKCUT_MAX_VIDEO_DURATION) || 60;
+var MAX_VIDEO_DIMENSION = Number(process.env.MARKCUT_MAX_VIDEO_DIMENSION) || 360;
+var DEFAULT_ITT_CLI = process.env.MARKCUT_ITT_CLI || 'uvx --from mlx-vlm mlx_vlm.generate --model mlx-community/MiniCPM-V-4.6-bf16 --max-tokens 2048 --prompt "{prompt}" --image {input} --temperature 0.0 --thinking-mode disabled';
+var GOOGLE_MAPS_API_KEY = process.env.GOOGLE_MAPS_API_KEY || "";
+var DEFAULT_VTT_SAMPLE_INTERVAL = Number(process.env.MARKCUT_VTT_SAMPLE_INTERVAL) || 5;
+var DEFAULT_VTT_CLI = process.env.MARKCUT_VTT_CLI || `uvx --from mlx-vlm mlx_vlm.generate --model mlx-community/MiniCPM-V-4.6-bf16 --max-tokens 2048 --prompt "{prompt}" --video {input} --temperature 0.0 --processor-kwargs '{"max_num_frames": 32, "stack_frames": 1, "max_slice_nums": 1, "use_image_id": false}'`;
 var DEFAULT_STT_CLI = process.env.MARKCUT_STT_CLI || 'uvx --from openai-whisper whisper "{input}" --output_format vtt --output_dir "{output}"';
-var DEFAULT_TTI_CLI = 'uvx --from mflux mflux-generate-flux2 --model flux2-klein-4b --steps 5 --prompt "{input}" --output "{output}"';
-var DEFAULT_TTV_CLI = "";
+var DEFAULT_TTS_CLI = process.env.MARKCUT_TTS_CLI || 'uvx edge-tts --voice "en-US-GuyNeural" --text "{input}" --write-media "{output}"';
+var DEFAULT_AGENT_CLI = process.env.MARKCUT_AGENT_CLI || "npx pi -p {prompt}";
+var DEFAULT_TTI_CLI = process.env.MARKCUT_TTI_CLI || 'uvx --from mflux mflux-generate-flux2 --model flux2-klein-4b --steps 5 --prompt "{input}" --output "{output}"';
+var DEFAULT_TTV_CLI = process.env.MARKCUT_TTV_CLI || "";
+
+// src/render/cli-tools.ts
 function substituteCli(template, input, output) {
   const safeOutput = output;
   let safeInput;
@@ -10128,7 +10118,8 @@ function preserveVariantAttrs(node2, attrs) {
     "transitionTime",
     "layout",
     "componentName",
-    "props"
+    "props",
+    "speaker"
   ]);
   for (const [k, v] of Object.entries(attrs)) {
     if (!STANDARD.has(k)) {
@@ -10225,6 +10216,7 @@ function parseNodeLine(content3, lineNum) {
         type: "audio",
         id: attrs.id,
         src,
+        speaker: attrs.speaker,
         duration: attrs.duration,
         start: attrs.start,
         startFrom: attrs.startFrom,
@@ -10302,12 +10294,12 @@ function parseNodeLine(content3, lineNum) {
     }
     case "script": {
       const raw = firstPositional ?? (attrs.script ? String(attrs.script) : void 0);
-      if (!raw) throw new DslError("script requires text content", ctx);
-      const text3 = isQuoted(raw) ? unquote(raw) : raw;
+      const text3 = raw ? isQuoted(raw) ? unquote(raw) : raw : void 0;
       const scriptNode = {
         type: "audio",
         id: attrs.id,
         script: text3,
+        speaker: attrs.speaker,
         volume: attrs.volume ?? 1,
         start: attrs.start,
         duration: attrs.duration,
@@ -10595,6 +10587,9 @@ function applyRootAttrs(root, attrs) {
         }
         break;
       }
+      case "voices":
+        root.voices = v;
+        break;
       default:
         throw new Error(`unknown root key: ${k}`);
     }
@@ -10735,29 +10730,92 @@ async function resolveMediaDurations(root, options = {}) {
   });
   return clone;
 }
+function parseDialogueLines(script) {
+  const lines = script.split("\n").map((l) => l.trim()).filter(Boolean);
+  const dialoguePattern = /^([\w\u4e00-\u9fff][\w\s\u4e00-\u9fff]*?):\s*(.+)$/;
+  const result = [];
+  for (const line of lines) {
+    const match = line.match(dialoguePattern);
+    if (!match) return [];
+    result.push({ speaker: match[1].trim(), text: match[2].trim() });
+  }
+  return result.length >= 2 ? result : [];
+}
+function resolveDialogue(root) {
+  const clone = JSON.parse(JSON.stringify(root));
+  const expansions = [];
+  walkDown(clone, (node2, parent) => {
+    if (node2.type !== "audio") return;
+    if (!node2.script || typeof node2.script !== "string") return;
+    if (node2.src) return;
+    const lines = parseDialogueLines(node2.script);
+    if (lines.length < 2) return;
+    if (!parent || !Array.isArray(parent.children)) return;
+    const idx = parent.children.indexOf(node2);
+    if (idx === -1) return;
+    expansions.push({ parent, index: idx, originalNode: node2, lines });
+  });
+  if (expansions.length === 0) return root;
+  expansions.sort((a, b) => b.index - a.index);
+  for (const { parent, index: index2, originalNode, lines } of expansions) {
+    const dialogueNodes = [];
+    for (let i = 0; i < lines.length; i++) {
+      const line = lines[i];
+      const { speaker, text: text3 } = line;
+      const lineNode = {
+        ...originalNode,
+        id: originalNode.id ? `${originalNode.id}-line-${i}` : void 0,
+        script: text3,
+        speaker,
+        start: 0,
+        duration: void 0
+        // Will be set by TTS duration probing
+      };
+      delete lineNode.src;
+      dialogueNodes.push(lineNode);
+    }
+    const seriesContainer = {
+      type: "series",
+      id: `${originalNode.id ?? "dialogue"}-series`,
+      children: dialogueNodes
+    };
+    parent.children.splice(index2, 1, seriesContainer);
+  }
+  const expanded = expansions.length > 0 ? clone : root;
+  const count = expansions.reduce((sum, e) => sum + e.lines.length, 0);
+  console.log(`  \u{1F4AC} dialogue: expanded ${expansions.length} script${expansions.length > 1 ? "s" : ""} into ${count} lines`);
+  return expanded;
+}
 async function resolveScripts(root, options) {
   const clone = JSON.parse(JSON.stringify(root));
   mkdirSync2(options.outputDir, { recursive: true });
   const cache = readCacheManifest(options.outputDir);
   let cacheDirty = false;
   const allScriptNodes = [];
-  walkDown(clone, (node2, parent) => {
+  walkDown(clone, (node2) => {
     if (node2.type !== "audio") return;
     if (!node2.script || typeof node2.script !== "string") return;
     if (node2.src) return;
     const id = node2.id ?? `audio-${allScriptNodes.length}`;
-    allScriptNodes.push({ node: node2, id, parent });
+    allScriptNodes.push({ node: node2, id });
   });
   if (allScriptNodes.length > 0) {
     console.log(`  \u{1F50A} TTS: generating ${allScriptNodes.length} script${allScriptNodes.length > 1 ? "s" : ""}...`);
   }
-  for (const { node: node2, id, parent } of allScriptNodes) {
-    const ttsCli = parent?.tts ?? clone.tts ?? options.ttsCli ?? DEFAULT_TTS_CLI;
+  for (const { node: node2, id } of allScriptNodes) {
+    let ttsCli = clone.tts ?? options.ttsCli ?? DEFAULT_TTS_CLI;
+    if (node2.speaker && clone.voices) {
+      const speakerVoice = clone.voices[node2.speaker];
+      if (speakerVoice) {
+        ttsCli = `${ttsCli} ${speakerVoice}`;
+      }
+    }
     const cacheKey = computeCacheKey({ script: node2.script, cli: ttsCli });
     const audioPath = join(options.outputDir, `${cacheKey}.mp3`);
     const cached = checkCache(cache, `tts:${cacheKey}`, cacheKey);
     let generated;
-    const label = firstWords(node2.script, 8);
+    const speakerLabel = node2.speaker ? `${node2.speaker}: ` : "";
+    const label = `${speakerLabel}${firstWords(node2.script, 8)}`;
     if (cached) {
       generated = cached;
       console.log(`  \u2713 TTS: ${label} (cached)`);
@@ -10799,10 +10857,10 @@ async function resolveSubtitles(root, options) {
     for (const node2 of nodes) {
       if (node2.isBackground) continue;
       const nodeStart = parentIsSeries ? seriesOffset : parentOffset;
-      const actionStart = node2.actions?.[0]?.start ?? 0;
+      const actionStart = node2.start ?? 0;
       const effectiveOffset = nodeStart + actionStart;
       if (node2.type === "audio" && node2.src) {
-        clips.push({ audioSrc: node2.src, offset: effectiveOffset });
+        clips.push({ audioSrc: node2.src, offset: effectiveOffset, speaker: node2.speaker });
       }
       if (node2.children && node2.children.length > 0) {
         let childIsSeries = false;
@@ -10838,7 +10896,7 @@ async function resolveSubtitles(root, options) {
   const mergedLines = ["WEBVTT", ""];
   let cueIndex = 1;
   let sttFailedCount = 0;
-  for (const { audioSrc, offset } of clips) {
+  for (const { audioSrc, offset, speaker } of clips) {
     const audioHash = existsSync2(audioSrc) ? createHash("sha1").update(readFileSync(audioSrc)).digest("hex").slice(0, 12) : audioSrc;
     const sttCacheKey = computeCacheKey({ audioHash, cli: sttCli });
     const sttKey = `stt:${audioSrc.split("/").pop()}`;
@@ -10883,7 +10941,10 @@ async function resolveSubtitles(root, options) {
         const sec = (s % 60).toFixed(3).padStart(6, "0");
         return `${String(h).padStart(2, "0")}:${String(m).padStart(2, "0")}:${sec}`;
       };
-      const text3 = lines.slice(lines.indexOf(tline) + 1).join("\n").trim();
+      let text3 = lines.slice(lines.indexOf(tline) + 1).join("\n").trim();
+      if (speaker) {
+        text3 = `${speaker}: ${text3}`;
+      }
       mergedLines.push(String(cueIndex++));
       mergedLines.push(`${formatSec(toSec(a) + offset)} --> ${formatSec(toSec(z) + offset)}`);
       mergedLines.push(text3, "");
@@ -11056,6 +11117,7 @@ async function resolveAll2(root, options = {}) {
     baseDir: options.baseDir,
     skip: options.skip
   });
+  result = resolveDialogue(result);
   if (options.scriptOutputDir) {
     result = await resolveScripts(result, {
       outputDir: options.scriptOutputDir,
@@ -11101,7 +11163,9 @@ async function resolveAndCompile(data, options = {}) {
     subtitleOutputDir: options.subtitleOutputDir,
     variants: options.variants
   });
-  const compiled = compileDescriptiveRoot(resolved);
+  const compiled = compileDescriptiveRoot(resolved, {
+    googleMapsApiKey: process.env.GOOGLE_MAPS_API_KEY
+  });
   return compiled;
 }
 async function resolveAndCompileMarkdown(markdown, options = {}) {

package/src/player/pipeline.ts

@@ -94,7 +94,9 @@ export async function resolveAndCompile(
   });
 
   // 2. Sync compile: descriptive → stream tree
-  const compiled = compileDescriptiveRoot(resolved);
+  const compiled = compileDescriptiveRoot(resolved, {
+    googleMapsApiKey: process.env.GOOGLE_MAPS_API_KEY,
+  });
 
   return compiled;
 }

package/src/player/server-shared.mjs

@@ -1,9 +1,8 @@
 #!/usr/bin/env node
 /**
- * Shared utilities for markcut player servers.
+ * Shared utilities for the markcut player server.
  *
- * Extracts common code between server.mjs (edit/preview) and
- * label-server.mjs (label annotation) into one place.
+ * Common code used by server.mjs across all modes (preview, edit, label).
  */
 
 import { statSync, createReadStream, readFileSync, existsSync } from "node:fs";
@@ -41,8 +40,8 @@ export function extractScenes(root) {
       if (!(s.type === "folder" || s.type === "scene" || s.children?.length)) continue;
       if (s.isBackground) continue;
       const leaf = (s.children || []).find(c => c.src && (c.type === "image" || c.type === "video"));
-      const action = leaf?.actions?.[0] || s.actions?.[0] || {};
-      const dur = (action.end || 5) - (action.start || 0);
+      const src2 = leaf || s;
+      const dur = (src2.end ?? 5) - (src2.start ?? 0);
       scenes.push({
         name: s.name || s.id || "scene",
         start: offset,
@@ -62,8 +61,7 @@ export function extractScenes(root) {
     let offset = 0;
     for (const s of scenesFolder.children) {
       const child = s.children?.[0] || {};
-      const action = child?.actions?.[0];
-      const dur = action ? (action.end - action.start) : 5;
+      const dur = (child.end ?? ((child.start ?? 0) + 5)) - (child.start ?? 0);
       scenes.push({
         name: s.name,
         start: offset,