@ignix-social/editor 0.1.22 → 0.1.24

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ignix-social/editor",
-  "version": "0.1.22",
+  "version": "0.1.24",
   "description": "Ignix Editor: render HTML video compositions (reels, edits, clips) on your own computer. Based on HyperFrames.",
   "license": "Apache-2.0",
   "type": "module",

package/skills/ignix-editor/SKILL.md

@@ -13,7 +13,7 @@ Commands. `ignix-editor` means the copy the Ignix playbook installed: `~/.ignix-
 |---|---|
 | `ignix-editor doctor` | Node 22+ and the H.264 encoder; downloads its own ffmpeg and whisper the first time. Run it first. |
 | `ignix-editor init <name>` | Creates `~/ignix-videos/<name>` with a working skeleton, GSAP, local fonts and `ignix-frame.js`. |
-| `ignix-editor transcribe <file> --lang es [--script "…"] -o words.json` | Words with start/end seconds, on this machine. |
+| `ignix-editor transcribe <file> --lang es [--script "…"] -o words.json` | Words with start/end seconds, plus `pauses` ([start, end] of every silence ≥0.35 s, for cutting silences), on this machine. Over 2 min it prints whisper's progress; without a GPU it takes about as long as the audio. |
 | `ignix-editor faces <video> -o work/faces.json` | Where the face is over time (macOS and Windows; the first run downloads the face detector): a smoothed path `{t, cx, cy, size}` (0–1, top-left origin) to frame the person. |
 | `ignix-editor pick-audio [dir] --kind music\|sfx --query "inspiring" [--order trending]` | Music (or a specific sound effect) from Pixabay, chosen by the USER: opens Pixabay's search in their browser, waits for their download, copies it into `assets/` and notes its source in `work/audio-sources.json`. Tell the user first that you opened Pixabay for them to pick one. Never fetch pixabay.com yourself: it blocks robots. |
 | `ignix-editor giphy --search "thumbs up" [--kind gifs\|stickers\|clips]` · `ignix-editor giphy <link> [dir]` | GIFs, stickers and clips from GIPHY, chosen by the USER: the first opens GIPHY's search in their browser; they copy the link of the one they like (Share → Copy link) and the second brings it into `assets/`: a GIF as MP4, a clip as MP4 with its audio, a sticker as a transparent frame sheet (see section 2, GIPHY stickers). |
@@ -114,7 +114,7 @@ Decide the zones once per video (e.g. header y 270–440, stage y 460–1300, ca
 
 Ready-made pieces that draw themselves from `t`, like everything else. They are options, not a format: use one when the idea is better told that way, never because it exists. `ignix-editor block` lists them; `ignix-editor block <name> <project>` copies `<name>.js` into the project; load it right after `ignix-frame.js`.
 
-**`stickman`** — stick-figure characters that walk, gesture, talk and react (explainers, "you vs. the smart way" sketches, a two-person dialogue, a story told by a character). Poses: `idle talk wave point think shrug cheer arms_crossed present sad sit jump_crouch jump_air`. Faces: `neutral happy sad surprised thinking angry`. Look: `color` (use the brand colour), `hair` (`short spiky curly long bun ponytail none`), `hairColor`, `accessory` (`none glasses cap headphones bow tie`), `theme` (`light`, or `dark` on a dark background).
+**`stickman`** — stick-figure characters that walk, gesture, talk and react (explainers, "you vs. the smart way" sketches, a two-person dialogue, a story told by a character). Poses: `idle talk wave point think shrug cheer arms_crossed present sad sit jump_crouch jump_air`. Faces: `neutral happy sad surprised thinking angry`. Look: `color` (use the brand colour), `hair` (`short spiky curly long bun ponytail none`), `hairColor`, `accessory` (`none glasses cap headphones bow tie`), `theme` (`light`, or `dark` on a dark background; pass `theme` in the state to `draw()` to switch it per frame, e.g. dark scene → white whiteboard, with the same figure).
 
 ```html
 <svg id="stage" viewBox="0 0 1080 1920" style="position:absolute;inset:0;width:1080px;height:1920px"></svg>
@@ -159,7 +159,7 @@ const chat = IgnixChat.create(document.getElementById('screen'), {
 run({ id, duration, fps }, (t) => chat.draw(t));
 ```
 
-**`whiteboard`** — a whiteboard explanation: shapes and words drawn by hand while a marker follows the stroke. Hand-drawn helpers (same wobble every frame): `IgnixBoard.line`, `rect`, `circle`, `arrow`, `underline`, `strike`; or any SVG path `d`. Each `shape(d, at, dur, { color, stroke, fill })` draws itself from `at` over `dur`; `write(text, x, y, at, dur, { size })` writes text left to right. One thing at a time, timed to the words, and big: on a 1080×1920 reel words of 70–110 px and shapes that use most of the width.
+**`whiteboard`** — a whiteboard explanation: shapes and words drawn by hand while a marker follows the stroke. Hand-drawn helpers (same wobble every frame): `IgnixBoard.line`, `rect`, `circle`, `arrow`, `underline`, `strike`; or any SVG path `d`. Each `shape(d, at, dur, { color, stroke, fill })` draws itself from `at` over `dur`; `write(text, x, y, at, dur, { size })` writes text left to right. `create(target, { paper })` also draws the board itself (`paper: '#F7F5F0'` or `{ color, x, y, w, h, radius, frame }`); `target` may be a `<g>` inside the `<svg>`, so scaling/translating that `<g>` moves a camera over the board. One thing at a time, timed to the words, and big: on a 1080×1920 reel words of 70–110 px and shapes that use most of the width.
 
 ```js
 const B = IgnixBoard;
@@ -259,7 +259,7 @@ work/      transcripts, edit decisions (JSON)
 build.mjs  reads work/ + durations (ignix-editor ffprobe) → writes assets/timing.js and the <video>/<audio> tags between markers in index.html
 ```
 
-Re-run it whenever the voice or the cuts change; the animation code reads `window.R` from `assets/timing.js` and never hardcodes seconds.
+Re-run it whenever the voice or the cuts change; the animation code reads `window.R` from `assets/timing.js` and never hardcodes seconds. Keep any HTML template the script reads in `work/`, never next to `index.html`: a second root HTML with `data-composition-id` fails `check` (`multiple_root_compositions`).
 
 ## 6. Traps that do not error
 
@@ -279,5 +279,5 @@ Re-run it whenever the voice or the cuts change; the animation code reads `windo
 Only if the app offers no way, give the user ONE line to paste into Terminal, e.g. `cd ~/ignix-videos/<name> && ~/.ignix-editor/cli/node_modules/.bin/ignix-editor render`, never the contents of a script file (a `#!` line breaks zsh), then continue when they say it finished.
 
 1. `ignix-editor check` → 0 errors.
-2. `ignix-editor snapshot --at <every ~0.5 s and every cue>` and LOOK at the frames (or tile them: `ignix-editor ffmpeg -pattern_type glob -i 'snapshots/frame-*.png' -vf "scale=270:480,tile=8x2" sheet.jpg`). Check: hook readable at frame 0, nothing still for >1.5 s, every named thing pointed at on its word, captions never over what is pointed at, nothing outside the safe zone.
+2. `ignix-editor snapshot --at <every ~0.5 s and every cue>` and LOOK at the frames — it also writes them tiled as `snapshots/contact-sheet*.jpg`, the quickest to read. Check: hook readable at frame 0, nothing still for >1.5 s, every named thing pointed at on its word, captions never over what is pointed at, nothing outside the safe zone.
 3. `ignix-editor render` → it reviews the MP4 by itself and prints `✓ Review: ready to show` or `✗ Review: NOT READY` with what to fix and where (stretches where nothing on the stage moves, a mostly empty frame, silences between voice lines, no change of scene). NOT READY means it is not finished: fix every ✗, render again, and only then show it. The `!` lines are worth fixing too when it is quick. (`ignix-editor review <mp4>` runs the same check on any video.) Then open the MP4's first frame and a few cues again, and deliver as the playbook says.

package/src/cli.mjs

@@ -12,7 +12,7 @@ import { HOME, PLATFORM, TOOLS, ensureTool, ensureWhisperModel, findTool } from
 import { linkMedia, prepare, projectHtml } from './prepare.mjs';
 import { validateHtml } from './validate.mjs';
 import { engineEnv, renderProject, runEngine } from './engine.mjs';
-import { alignToScript, transcribe } from './transcribe.mjs';
+import { alignToScript, fitToPauses, transcribe } from './transcribe.mjs';
 import { faces as detectFaces } from './analysis/detect.mjs';
 import { frames, probe } from './analysis/media.mjs';
 import { serve } from './analysis/serve.mjs';
@@ -215,16 +215,21 @@ const commands = {
     const ffprobe = await ensureTool('ffprobe', log);
     const whisper = await ensureTool('whisper', log);
     const model = await ensureWhisperModel(opts.model ?? 'medium', log);
-    let words = transcribe({ ffmpeg, whisper, model, input, language: opts.lang ?? 'es', threads: opts.threads });
-    if (opts.script) {
-      const dur = Number(spawnSync(ffprobe, ['-v', 'error', '-show_entries', 'format=duration', '-of', 'csv=p=0', input], { encoding: 'utf8' }).stdout.trim());
-      words = alignToScript(opts.script, words, dur);
-    }
-    const json = JSON.stringify({ language: opts.lang ?? 'es', words }, null, 1);
+    const dur = Number(spawnSync(ffprobe, ['-v', 'error', '-show_entries', 'format=duration', '-of', 'csv=p=0', input], { encoding: 'utf8' }).stdout.trim()) || 0;
+    const mmss = (x) => `${Math.floor(x / 60)}:${String(Math.round(x % 60)).padStart(2, '0')}`;
+    const t0 = Date.now();
+    // Aviso de duración ANTES de empezar: sin GPU, una charla larga tarda más o menos lo que dura.
+    if (dur > 120) log(`Transcribing ${mmss(dur)} of audio with the ${basename(model).replace(/^ggml-|\.bin$/g, '')} model. Without a GPU this can take about as long as the audio itself; whisper prints its progress below.`);
+    const r = transcribe({ ffmpeg, whisper, model, input, language: opts.lang ?? 'es', threads: opts.threads, progress: dur > 120 });
+    let words = r.words;
+    if (opts.script) words = alignToScript(opts.script, words, dur);
+    words = fitToPauses(words, r.pauses);
+    const json = JSON.stringify({ language: opts.lang ?? 'es', duration: +dur.toFixed(3), words, pauses: r.pauses }, null, 1);
     if (opts.output) {
       mkdirSync(dirname(resolve(opts.output)), { recursive: true });
       writeFileSync(resolve(opts.output), json);
     } else console.log(json);
+    log(`✓ ${words.length} words and ${r.pauses.length} pauses (≥0.35 s) in ${mmss((Date.now() - t0) / 1000)}${opts.output ? ` → ${opts.output}` : ''}`);
   },
 
   // Música o efectos de Pixabay elegidos por la persona (src/pick-audio.mjs): abre la
@@ -296,7 +301,7 @@ const commands = {
     if (r.sheet) {
       const sh = r.sheet;
       log(`✓ sticker → ${r.file}: ${sh.frames} frames in a ${sh.cols}×${sh.rows} transparent sheet, ${sh.frameWidth}×${sh.frameHeight} each, ${sh.fps} fps (${sh.durationSec} s loop). Animate it from t as the skill shows (GIPHY section).`);
-    } else log(`✓ ${r.kind} → ${r.file}${r.hasAudio ? ' (with its audio)' : ''}`);
+    } else log(`✓ ${r.kind} → ${r.file}${r.kind === 'clip' ? (r.hasAudio ? ' (with its audio)' : ' (this clip has no sound: treat it as a muted clip)') : ''}`);
     if (opts.json) console.log(JSON.stringify(r));
   },
 

package/src/giphy.mjs

@@ -67,7 +67,8 @@ export function giphySearchUrl(query, kind = 'gifs') {
 
 /** Qué fichero traer de media.giphy.com para cada tipo (y el de reserva). */
 export function giphyRenditions(kind) {
-  if (kind === 'clip') return [['giphy480p.mp4', 'mp4'], ['giphy.mp4', 'mp4']];
+  // El de 360p es el que lleva el audio; giphy480p.mp4 dejó de servirse (404, 2026-10-08).
+  if (kind === 'clip') return [['giphy360p.mp4', 'mp4'], ['giphy.mp4', 'mp4']];
   if (kind === 'sticker') return [['giphy.gif', 'gif']];
   return [['giphy.mp4', 'mp4'], ['giphy.gif', 'gif']];
 }
@@ -93,21 +94,21 @@ export function gifToSheet({ ffmpeg, ffprobe, gif, out }) {
 
 /**
  * El tipo cuando el enlace no lo dice (el del ARCHIVO, …/giphy.gif): con él un clip se traía
- * como GIF, sin audio. Solo los clips tienen giphy480p.mp4, y la página dice si es sticker.
+ * como GIF, sin audio. La página lo dice: los clips traen su bloque "video" y los stickers
+ * is_sticker.
  */
 export async function resolveGiphyKind(id, fromLink) {
   if (fromLink) return fromLink;
   if (!ID.test(id)) return 'gif';
-  const clip = await fetch(`https://media.giphy.com/media/${id}/giphy480p.mp4`, { method: 'HEAD', signal: AbortSignal.timeout(8000) })
-    .then((r) => r.ok)
-    .catch(() => false);
-  if (clip) return 'clip';
   const page = await fetch(`https://giphy.com/gifs/${id}`, { headers: { 'user-agent': 'Mozilla/5.0 (compatible; IgnixEditor)' }, signal: AbortSignal.timeout(8000) })
     .then((r) => (r.ok ? r.text() : ''))
     .catch(() => '');
   const html = page.replace(/\\"/g, '"');
   const at = html.indexOf(`"id":"${id}"`);
-  return at >= 0 && /"is_sticker":true/.test(html.slice(at, at + 8000)) ? 'sticker' : 'gif';
+  if (at < 0) return 'gif';
+  const next = html.slice(at + 10).search(/"id":"[A-Za-z0-9]{8,40}"/);
+  const own = html.slice(at, next < 0 ? at + 8000 : at + 10 + next);
+  return /"video":\{/.test(own) ? 'clip' : /"is_sticker":true/.test(own) ? 'sticker' : 'gif';
 }
 
 /** Trae el elegido a `<project>/assets/` y apunta su origen en `work/media-sources.json`. */
@@ -143,7 +144,12 @@ export async function importGiphy({ link, project, kind: forced, ffmpeg, ffprobe
       /* fichero roto: se empieza de nuevo */
     }
     writeFileSync(log, JSON.stringify([...prev, { source: 'giphy', kind, id: parsed.id, link: String(link), file: dest, at: new Date().toISOString() }], null, 2));
-    return { id: parsed.id, kind, file: dest, hasAudio: kind === 'clip' && file.includes('480p'), ...(sheet ? { sheet } : {}) };
+    // Audio de verdad, no supuesto: hay clips mudos («Take My Money»).
+    const hasAudio =
+      kind === 'clip' && ffprobe
+        ? spawnSync(ffprobe, ['-v', 'error', '-select_streams', 'a', '-show_entries', 'stream=index', '-of', 'csv=p=0', dest], { encoding: 'utf8' }).stdout.trim() !== ''
+        : false;
+    return { id: parsed.id, kind, file: dest, hasAudio, ...(sheet ? { sheet } : {}) };
   }
   throw new Error(`Could not download that ${kind} from GIPHY${lastErr ? `: ${lastErr.message}` : ''}.`);
 }

package/src/runtime/blocks/stickman.js

@@ -182,10 +182,16 @@
   // look: { color, hair: short|spiky|curly|long|bun|ponytail|none, hairColor, accessory: none|glasses|cap|headphones|bow|tie, theme: 'light'|'dark' }
   function create(svg, look = {}) {
     const L = { color: '#D24D1E', hair: 'short', hairColor: '#22201E', accessory: 'none', theme: 'light', ...look };
-    const dark = L.theme === 'dark';
-    const ink = dark ? '#F3F0EA' : '#1F1D1B';
-    const headFill = dark ? '#2A3038' : '#FFFFFF';
-    const hairFill = dark && lum(L.hairColor) <= 0.25 ? '#D9D4CC' : L.hairColor;
+    // Colores según el tema; `draw` puede cambiar de tema en cada fotograma (escena oscura →
+    // pizarra blanca) sin crear otro monigote.
+    let ink, headFill, hairFill;
+    function palette(theme) {
+      const dark = theme === 'dark';
+      ink = dark ? '#F3F0EA' : '#1F1D1B';
+      headFill = dark ? '#2A3038' : '#FFFFFF';
+      hairFill = dark && lum(L.hairColor) <= 0.25 ? '#D9D4CC' : L.hairColor;
+    }
+    palette(L.theme);
     const tie = lum(L.color) > 0.2 ? '#2B2F3A' : '#E8E2D8';
     const g = document.createElementNS('http://www.w3.org/2000/svg', 'g');
     svg.appendChild(g);
@@ -263,9 +269,10 @@
       return s;
     }
 
-    // state: { x, ground, scale, facing, joints, face, mouth, eyesOpen, alpha, pop, lift }
+    // state: { x, ground, scale, facing, joints, face, mouth, eyesOpen, alpha, pop, lift, theme }
     function draw(state) {
-      const st = { scale: 1, facing: 1, mouth: 0, eyesOpen: 1, alpha: 1, pop: 1, lift: 0, face: 'neutral', joints: 'idle', ...state };
+      const st = { scale: 1, facing: 1, mouth: 0, eyesOpen: 1, alpha: 1, pop: 1, lift: 0, face: 'neutral', joints: 'idle', theme: L.theme, ...state };
+      palette(st.theme);
       const joints = typeof st.joints === 'string' ? POSES[st.joints] ?? POSES.idle : { ...DEFAULT, ...st.joints };
       const face = typeof st.face === 'string' ? EXPRESSIONS[st.face] ?? EXPRESSIONS.neutral : { ...EXPRESSIONS.neutral, ...st.face };
       const p = forward(joints);

package/src/runtime/blocks/whiteboard.js

@@ -57,8 +57,19 @@
   const underline = (x, y, w, seed = seedN++) => line(x, y, x + w, y + 6, 5, seed);
   const strike = (x, y, w, seed = seedN++) => `${line(x, y, x + w, y - 10, 4, seed)} ${line(x + w * 0.05, y + 14, x + w * 1.02, y + 2, 4, seed + 1)}`;
 
-  function create(svg, { ink = '#1F1D1B', width = 9, marker = true, markerColor = '#2B6CB0' } = {}) {
+  // `svg` puede ser el <svg> o un <g> dentro de él (para mover una «cámara» sobre la
+  // pizarra: escala/traslada el <g>). `paper` dibuja la pizarra en sí: un color, o
+  // { color, x, y, w, h, radius, frame } (por defecto 0,0,1080,1920 sin marco).
+  function create(svg, { ink = '#1F1D1B', width = 9, marker = true, markerColor = '#2B6CB0', paper = null } = {}) {
     const items = [];
+    if (paper) {
+      const P = typeof paper === 'string' ? { color: paper } : paper;
+      const bg = document.createElementNS(NS, 'rect');
+      const attrs = { x: P.x ?? 0, y: P.y ?? 0, width: P.w ?? 1080, height: P.h ?? 1920, rx: P.radius ?? 0, fill: P.color ?? '#F7F5F0' };
+      if (P.frame) Object.assign(attrs, { stroke: P.frame === true ? '#C9C4BA' : P.frame, 'stroke-width': P.frameWidth ?? 14 });
+      for (const [k, v] of Object.entries(attrs)) bg.setAttribute(k, String(v));
+      svg.appendChild(bg);
+    }
     const layer = document.createElementNS(NS, 'g');
     svg.appendChild(layer);
     let pen = null;
@@ -109,11 +120,12 @@
       tx.style.font = font;
       tx.style.fontSize = `${size}px`;
       layer.appendChild(tx);
-      // Ancho estimado sin medir, por lo alto (~0,8 em por carácter en una negrita): la
-      // máscara solo descubre, así que pasarse no corta nada; al acabar se quita.
-      const w = text.length * size * 0.8;
-      const x0 = anchor === 'middle' ? x - w / 2 : anchor === 'end' ? x - w : x;
-      items.push({ kind: 'text', el: tx, rect: r, at, dur, x0, w, y: y - size, h: size * 1.4 });
+      // El ancho se MIDE en cada fotograma (getComputedTextLength: geometría del texto con
+      // la fuente ya cargada, igual en todos los fotogramas del render) para que el
+      // rotulador acabe donde acaba la palabra. Si aún no se puede medir, estimación por lo
+      // alto (~0,8 em por carácter): la máscara solo descubre y al acabar se quita.
+      const est = text.length * size * 0.8;
+      items.push({ kind: 'text', el: tx, rect: r, at, dur, x, anchor, est, y: y - size, h: size * 1.4 });
       return tx;
     }
 
@@ -127,6 +139,14 @@
           if (it.fillEl) it.fillEl.style.opacity = String(clamp01((t - it.at - it.dur) / 0.3));
           if (k > 0 && k < 1) tip = it.el.getPointAtLength(it.len * k);
         } else {
+          let w = it.est;
+          try {
+            w = it.el.getComputedTextLength() || it.est;
+          } catch {
+            /* sin medida: estimación */
+          }
+          it.w = w;
+          it.x0 = it.anchor === 'middle' ? it.x - w / 2 : it.anchor === 'end' ? it.x - w : it.x;
           it.rect.setAttribute('y', String(it.y));
           it.rect.setAttribute('height', String(it.h));
           // Terminado, sin máscara: que ninguna letra se quede fuera por la estimación.

package/src/transcribe.mjs

@@ -46,19 +46,60 @@ function whisperTimeoutMs(wavPath) {
   } catch {
     /* sin tamaño: el mínimo */
   }
-  return Math.max(10 * 60_000, seconds * 6 * 1000);
+  // ENTERO: spawnSync rechaza un timeout con decimales («The value of "timeout" is out of
+  // range… Received 13158197.625») y eso rompía cualquier audio de más de 100 s (2026-10-08).
+  return Math.ceil(Math.max(10 * 60_000, seconds * 6 * 1000));
+}
+
+/**
+ * Pausas del audio (ffmpeg silencedetect): [[inicio, fin], …] en segundos. Sirven para el
+ * paso de silencios y para que ninguna palabra se estire sobre una pausa: whisper con -ml 1
+ * hace que cada palabra acabe donde empieza la siguiente («claro.» de 0,76 s, «1.600,» de 1,6 s).
+ */
+export function detectPauses(ffmpeg, input, { noiseDb = -35, minSec = 0.35 } = {}) {
+  const r = spawnSync(ffmpeg, ['-nostdin', '-hide_banner', '-i', input, '-vn', '-af', `silencedetect=noise=${noiseDb}dB:d=${minSec}`, '-f', 'null', '-'], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024 });
+  const out = [];
+  let start = null;
+  for (const line of (r.stderr || '').split('\n')) {
+    const a = line.match(/silence_start: (-?[\d.]+)/);
+    if (a) start = Math.max(0, Number(a[1]));
+    const b = line.match(/silence_end: ([\d.]+)/);
+    if (b && start !== null) {
+      out.push([+start.toFixed(3), +Number(b[1]).toFixed(3)]);
+      start = null;
+    }
+  }
+  return out;
+}
+
+/** Recorta cada palabra a la voz: ni empieza ni acaba dentro de una pausa (si le queda voz). */
+export function fitToPauses(words, pauses) {
+  if (!pauses?.length) return words;
+  return words.map((w) => {
+    let { s, e } = w;
+    for (const [ps, pe] of pauses) {
+      if (ps >= e || pe <= s) continue;
+      if (ps > s && ps - s >= 0.08) e = Math.min(e, ps); // se estiraba sobre la pausa
+      else if (pe < e && e - pe >= 0.08) s = Math.max(s, pe); // empezaba dentro de la pausa
+    }
+    return { ...w, s: +s.toFixed(3), e: +e.toFixed(3) };
+  });
 }
 
 function whisperFailure(w, timeoutMs) {
   if (w.error?.code === 'ETIMEDOUT') {
     return new Error(`whisper did not finish in ${Math.round(timeoutMs / 60_000)} min and was stopped. Close other heavy programs and run the command again, or use a smaller model (--model small).`);
   }
-  return new Error(`whisper failed: ${(w.stderr || '').slice(-400)}`);
+  return new Error(`whisper failed (exit ${w.status ?? w.error?.code ?? '?'}): ${(w.stderr || '').slice(-400)}`);
 }
 
 const norm = (s) => s.toLowerCase().normalize('NFD').replace(/[̀-ͯ]/g, '').replace(/[^\p{L}\p{N}]/gu, '');
 
-export function transcribe({ ffmpeg, whisper, model, input, language = 'es', threads }) {
+/**
+ * Palabras y pausas. Con `progress`, whisper escribe su avance («progress = 40%») en la
+ * consola: en una transcripción larga, sin eso nadie distingue «va lento» de «se ha colgado».
+ */
+export function transcribe({ ffmpeg, whisper, model, input, language = 'es', threads, progress = false }) {
   const ws = whisperWorkspace(model);
   const tmp = ws.tmp;
   try {
@@ -67,8 +108,10 @@ export function transcribe({ ffmpeg, whisper, model, input, language = 'es', thr
     if (a.status !== 0) throw new Error(`Could not read the audio of ${input}: ${a.stderr.slice(-300)}`);
     const out = join(tmp, 'out');
     const args = ['-m', ws.modelArg, '-l', language, '-ml', '1', '-sow', '-oj', '-of', ws.arg(out), '-t', String(whisperThreads(threads)), '-f', ws.arg(wav)];
+    if (progress) args.unshift('-np', '-pp');
     const timeout = whisperTimeoutMs(wav);
-    const run = (extra) => spawnSync(whisper, [...extra, ...args], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024, cwd: ws.cwd, timeout });
+    const stdio = progress ? ['ignore', 'pipe', 'inherit'] : undefined;
+    const run = (extra) => spawnSync(whisper, [...extra, ...args], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024, cwd: ws.cwd, timeout, stdio });
     // En Mac Intel la GPU (Metal) se cuelga con whisper: CPU directamente. En el resto, GPU y,
     // si falla, otra vez por CPU.
     const intelMac = process.platform === 'darwin' && process.arch === 'x64';
@@ -82,7 +125,7 @@ export function transcribe({ ffmpeg, whisper, model, input, language = 'es', thr
       if (!t || /^\[.*\]$/.test(t)) continue; // [MÚSICA], [MUSIC]…
       words.push({ t, s: seg.offsets.from / 1000, e: seg.offsets.to / 1000 });
     }
-    return words;
+    return { words, pauses: detectPauses(ffmpeg, wav) };
   } finally {
     rmSync(tmp, { recursive: true, force: true });
   }
@@ -95,7 +138,7 @@ export function alignToScript(script, tokens, end) {
   let i = 0;
   for (let j = 0; j < tokens.length && i < sw.length; j++) {
     const tk = tokens[j];
-    const tEnd = j + 1 < tokens.length ? tokens[j + 1].s : Math.max(tk.e, end);
+    const tEnd = j + 1 < tokens.length ? tokens[j + 1].s : end > 0 ? end : tk.e;
     // whisper juntó dos palabras en una («a DM» → «ADM»)
     if (i + 1 < sw.length && norm(sw[i] + sw[i + 1]) === norm(tk.t) && norm(sw[i]) !== norm(tk.t)) {
       const mid = tk.s + (tEnd - tk.s) * (sw[i].length / (sw[i].length + sw[i + 1].length));
@@ -118,7 +161,10 @@ export function alignToScript(script, tokens, end) {
       s = e;
     }
   }
-  return out.map((w) => ({ t: w.t, s: +w.s.toFixed(3), e: +w.e.toFixed(3) }));
+  // whisper trabaja en ventanas de 30 s y a veces pone el final de la última palabra más
+  // allá del audio (una frase de 2,76 s con «Ignix» acabando en 3,6 s): nada pasa del final.
+  const cap = end > 0 ? (x) => Math.min(x, end) : (x) => x;
+  return out.map((w) => ({ t: w.t, s: +cap(w.s).toFixed(3), e: +cap(w.e).toFixed(3) }));
 }
 
 /**