@keepylabs/cutroom 0.5.2 → 0.5.4

package/cutroom.mjs

@@ -21894,6 +21894,10 @@ function cutAndNormalize(src, out, segments) {
   if (r.status !== 0) throw new Error(`Couldn't cut the footage:
 ${tail(r.stderr)}`);
 }
+function looksLikeSpeech(words, duration3) {
+  const readable = words.filter((w) => new RegExp("\\p{L}{2,}|\\d", "u").test(w.w) && !/\uFFFD/.test(w.w));
+  return readable.length >= Math.max(3, duration3 * 0.6) && readable.length >= words.length * 0.8;
+}
 async function prepareFootage(input) {
   const source = resolve2(input.video.replace(/^~/, homedir2()));
   if (!existsSync2(source)) throw new Error(`No such file: ${source}`);
@@ -21909,7 +21913,8 @@ async function prepareFootage(input) {
   const footage = join3(work, "footage.mp4");
   cutAndNormalize(source, footage, segments);
   const duration3 = probe(footage).duration;
-  const t = sh("npx", ["--yes", HF, "transcribe", "footage.mp4", "--json", "-d", "."], work);
+  const lang = (input.language ?? "en").toLowerCase().slice(0, 5);
+  const t = sh("npx", ["--yes", HF, "transcribe", "footage.mp4", "--json", "-d", ".", ...lang === "en" ? [] : ["--language", lang, "-m", "large-v3"]], work);
   const meta2 = (() => {
     try {
       return JSON.parse(t.stdout.trim().split("\n").pop() ?? "{}");
@@ -21921,6 +21926,7 @@ async function prepareFootage(input) {
 ${meta2.error ?? tail(t.stderr)}`);
   const raw = JSON.parse(readFileSync3(meta2.transcriptPath ?? join3(work, "transcript.json"), "utf8"));
   const words = raw.map((w) => ({ t: Math.round(w.start * 100) / 100, e: Math.round(w.end * 100) / 100, w: w.text.trim() })).filter((w) => w.w);
+  if (!looksLikeSpeech(words, duration3)) throw new Error(lang === "en" ? "I couldn't hear anyone talking in this video (it sounds like music or silence). Cutroom finishing needs you talking to camera. If you speak another language, tell me which one and I'll try again." : `I couldn't make out the speech in this video, even in that language. Check the sound is you talking to camera.`);
   const project = { name, folder, source, footage, duration: Math.round(duration3 * 100) / 100, words, removed_s: Math.round((p.duration - duration3) * 10) / 10, segments };
   writeFileSync2(projectFile(folder), JSON.stringify(project, null, 2));
   return {
@@ -21928,7 +21934,7 @@ ${meta2.error ?? tail(t.stderr)}`);
     duration: project.duration,
     pauses_removed_s: project.removed_s,
     transcript: words.map((w) => `${w.t.toFixed(2)} ${w.w}`).join("\n"),
-    next: "Check the transcript for misheard words (pass them as fixes). Get the menu with list_styles, pick a layout, a caption style and a few effects with start times from this transcript, then call render_reel."
+    next: "Check the transcript for misheard words (pass them as fixes). Then plan: list_styles, get_method('edit_plan') and follow it, then plan_reel to show the storyboard. Nothing renders until the user says go."
   };
 }
 async function ensureFonts(work) {
@@ -22061,18 +22067,27 @@ async function composition(input) {
   return { pj, work, html, cutout, words };
 }
 var fmtT = (t) => `${t.toFixed(1)}s`;
-var fxLine = (f) => Object.entries(f).filter(([k]) => !["slug", "start", "dur"].includes(k)).map(([, v]) => Array.isArray(v) ? v.join(" / ") : String(v)).join(" \xB7 ");
+var SHOWN = ["word", "text", "title", "number", "label", "eyebrow", "keyword", "items", "lines", "labels", "steps", "message", "before", "after", "note"];
+var fxLine = (f) => SHOWN.filter((k) => f[k] !== void 0 && f[k] !== "").map((k) => k === "keyword" ? `comment "${f[k]}"` : Array.isArray(f[k]) ? f[k].join(" / ") : String(f[k])).join(" \xB7 ");
 async function planReel(input) {
   const { pj, work, html, cutout, words } = await composition(input);
   const effects = [...input.effects ?? []].sort((a, b) => Number(a.start) - Number(b.start));
   const peak = (f) => {
     const d = Math.min(Math.max(Number(f.dur) || 2.5, 1.2), 5);
-    return Math.min(pj.duration - 0.1, Number(f.start) + Math.max(d * 0.5, d - 0.55));
+    const at = f.slug === "frame-rip" ? d * 0.5 : Math.max(d * 0.5, d - (f.slug === "comment-dm" ? 0.32 : 0.55));
+    return Math.min(pj.duration - 0.1, Number(f.start) + at);
   };
   const plainAt = effects.length && Number(effects[0].start) > 2 ? 1.2 : null;
+  const styles = await call("styles");
+  const title = (list, slug3) => (list.find((x) => x.slug === slug3)?.title ?? String(slug3)).replace(/\s*\(.*\)$/, "");
+  const said = (from, to) => words.filter((w) => w.t >= from - 0.05 && w.t < to).map((w) => w.w).join(" ");
+  const ends = effects.map((f, k) => Math.min(effects[k + 1] ? Number(effects[k + 1].start) : pj.duration, Number(f.start) + Math.min(Math.max(Number(f.dur) || 2.5, 1.2), 5)));
   const moments = [
-    ...plainAt !== null ? [{ t: plainAt, what: `${input.layout ?? "full-frame"} + ${input.caption ?? "two-tier"} captions`, tag: "layout + captions", at: plainAt }] : [],
-    ...effects.map((f) => ({ t: Number(f.start), what: `${f.slug}${fxLine(f) ? `: ${fxLine(f)}` : ""}`, tag: String(f.slug), at: peak(f) }))
+    ...plainAt !== null ? [{ t: plainAt, end: Number(effects[0].start), what: "just you and the captions", tag: "you + captions", at: plainAt }] : [],
+    ...effects.map((f, k) => {
+      const name = title(styles.effects, f.slug);
+      return { t: Number(f.start), end: ends[k], what: `${name}${fxLine(f) ? `: ${fxLine(f)}` : ""}`, tag: name, at: peak(f) };
+    })
   ];
   const shots = join3(work, "storyboard");
   rmSync(shots, { recursive: true, force: true });
@@ -22100,11 +22115,15 @@ ${tail(t.stderr)}`);
   const plan = { layout: input.layout ?? "full-frame", caption: input.caption ?? "two-tier", effects, images: input.images, emphasis: input.emphasis, fixes: input.fixes };
   writeFileSync2(planFile(pj.folder), JSON.stringify(plan, null, 2));
   const text = [
-    `PLAN  ${pj.name}  ${pj.duration.toFixed(1)}s  ${effects.length} effects  layout: ${plan.layout}  captions: ${plan.caption}${cutout ? "  (cut-out ready)" : ""}`,
-    ...moments.map((m) => `${fmtT(m.t).padStart(6)}  ${m.what}`),
+    `Plan for ${pj.name} (${pj.duration.toFixed(1)}s). Layout: ${title(styles.layouts, plan.layout)}. Captions: ${title(styles.captions, plan.caption)}.${cutout ? " You are cut out of the background for this one (done)." : ""}`,
+    "",
+    "| Time | What you say | What shows on screen |",
+    "| - | - | - |",
+    ...moments.map((m) => `| ${m.t.toFixed(1)} to ${m.end.toFixed(1)}s | ${said(m.t, m.end).replace(/\|/g, "/") || "(pause)"} | ${m.what.replace(/\|/g, "/")} |`),
+    "",
     `Storyboard: ${board}`
   ].join("\n");
-  return { text, storyboard: board, plan, words_used: words.length, next: "Show the storyboard. On changes, plan_reel again with the new choice. On a go, render_reel with plan:true and draft:true." };
+  return { text, storyboard: board, plan, words_used: words.length, next: "Show this table as it is (plain words, no effect codes) and the storyboard. On changes, plan_reel again with the new choice. On a go, render_reel with plan:true and draft:true." };
 }
 async function renderReel(input) {
   if (input.plan) {
@@ -22128,7 +22147,7 @@ ${tail(r.stderr) || tail(r.stdout)}`);
   renameSync(tmp, out);
   const render_s = Math.round((Date.now() - started) / 100) / 10;
   const qa = await qaVideo({ video: out, write_report: true, layout: input.layout });
-  return { video: out, render_s, ...cutout?.made_s ? { cutout_s: cutout.made_s } : {}, qa, next: qa.passed ? "Watch it once, then it's ready to post." : "Fix the issues listed (usually: move an effect, fix a word, or re-record a bit), then render_reel again." };
+  return { video: out, render_s, ...cutout?.made_s ? { cutout_s: cutout.made_s } : {}, qa, next: qa.passed ? "Watch it once, then it's ready to post." : qa.technical?.length ? "Fix the technical issues listed, then render again." : "Look at each flagged spot yourself (grab the frame) and tell the user which are real problems, with the fix, and which are the design." };
 }
 async function qaVideo(input) {
   const file = resolve2(input.video.replace(/^~/, homedir2()));
@@ -22179,29 +22198,23 @@ process.env.PATH = [process.env.PATH, "/opt/homebrew/bin", "/usr/local/bin", joi
 var result = (value) => ({ content: [{ type: "text", text: typeof value === "string" ? value : JSON.stringify(value, null, 2) }] });
 var failure2 = (error2) => ({ isError: true, content: [{ type: "text", text: error2 instanceof Error ? error2.message : String(error2) }] });
 var safe = (op) => Promise.resolve().then(op).then(result).catch(failure2);
-var INSTRUCTIONS = `Cutroom: short-form video help for the user. Speak plainly; assume they are not technical.
-
-Setup or any failure: call cutroom_doctor and show a short checklist. Missing ffmpeg: ask, then run "brew install ffmpeg". Missing Node.js: ask, then run "brew install node". Key problem: tell them to reconnect with the one install command from their welcome email; never ask them to paste the key into the chat.
+var INSTRUCTIONS = `Cutroom: short-form video help. Speak plainly; the user is not technical. Never show tool or effect codes; use plain names.
 
-Hooks, scripts, captions: call get_method ('hooks', 'script' or 'caption') and follow it step by step, scoring with score_hooks or score_script as the method says.
+Setup or any failure: cutroom_doctor, then a short checklist. Missing ffmpeg or Node.js: ask, then run "brew install ffmpeg" or "brew install node". Key problem: reconnect Cutroom with their key from the Cutroom key space on Whop; never ask them to paste the key into the chat.
 
-Break down a reel (link or file): call breakdown_reel, then get_method('breakdown_writeup') and follow it.
+Finish a video (a raw video of them talking: edit it, captions, effects, make it a reel): get_method('finish_video') first and follow it step by step. Nothing renders until they approve the plan.
 
-Tear down a creator: call creator_teardown (compare_to = the user's own handle if known), then get_method('teardown_writeup').
+Hooks, scripts, captions: get_method('hooks', 'script' or 'caption') and follow it.
+Break down a reel: breakdown_reel, then get_method('breakdown_writeup').
+Tear down a creator: creator_teardown (compare_to = their handle), then get_method('teardown_writeup').
 
-Finish a video (raw talking-head file: edit it, add captions or effects, make it a reel):
-1. prepare_footage with the file path.
-2. From the timed transcript, list words that look misheard (names, brands) and confirm spellings with the user.
-3. Plan: list_styles (share its gallery link so the user can see each style moving), then choose one layout, one caption style and 2 to 4 effects tied to moments in the transcript (a number gets number-pop, steps get checklist, the CTA line gets comment-dm or follow-tap). Effect text uses only the user's own words and numbers. Call plan_reel with that choice and show the user the plan lines and the storyboard (one frame per moment). They change anything ("move the checklist to 30s", "swap orbit for card wall"): plan_reel again. Nothing renders until they say go. No B-roll unless they ask for it. Effects and layouts marked needs "cutout" (text behind you, orbit, card wall, cut-out layouts) first cut the speaker out of the background on this computer: about 6 seconds of work per second of video, once per video; say so before you start. Ones marked needs "images" take the user's own pictures as file paths; ask for them, or use the fallback the effect describes.
-4. On a go: render_reel with plan:true and draft:true; tell them where the file is and what QA found.
-5. After they approve the draft: render_reel with plan:true (no draft) for the final. After changes: plan_reel, then render again.
-6. QA "look" lines are spots to check, not failures: busy edits often get flagged on purpose (cards running off the edge, a card over a board). For each flagged time, grab that frame with ffmpeg and look at it yourself, then tell the user which ones are real problems (with the fix) and which are the design. Technical issues (frozen picture, loudness, black frames) are real; say if a freeze is a deliberate hold.
+Before a first render: show_brand; if none, offer set_brand (from colors, a website or a logo: pick background, accent, highlight, show them, save on a yes).
 
-Brand: before a user's first render, call show_brand; if none is set, offer to set it with set_brand. If they give a website or logo instead of colors, pick 3 colors from it (background, accent, highlight), show them, and save on a yes.
+QA flags are spots to check: look at each flagged frame yourself and say which are real problems (with the fix) and which are the design.
 
-Results are saved in ~/Cutroom. Study only for other creators' reels: never reuse their words, footage or branding.`;
+Results are saved in ~/Cutroom. Other creators' reels are for study only: never reuse their words, footage or branding.`;
 await serveStdio(() => {
-  const server = new McpServer({ name: "cutroom", version: "0.5.2" }, { instructions: INSTRUCTIONS });
+  const server = new McpServer({ name: "cutroom", version: "0.5.4" }, { instructions: INSTRUCTIONS });
   server.registerTool("cutroom_doctor", {
     description: "Check Cutroom is ready: ffmpeg on this computer and an active membership key. Returns each check with the exact fix. Run it first, and whenever a tool fails.",
     inputSchema: object({})
@@ -22260,7 +22273,8 @@ ${m.text}`;
     inputSchema: object({
       video: string2().min(1).describe("Path to the raw video file on this computer."),
       name: string2().optional().describe("Short name for the output, e.g. 'trial-loop'."),
-      cut_pauses: boolean2().default(true).describe("Cut pauses longer than about half a second.")
+      cut_pauses: boolean2().default(true).describe("Cut pauses longer than about half a second."),
+      language: string2().optional().describe("Spoken language code if not English, e.g. ko, es, ja (uses a larger model, downloaded once).")
     })
   }, async (input) => safe(() => prepareFootage(input)));
   server.registerTool("plan_reel", {
@@ -22269,7 +22283,7 @@ ${m.text}`;
       project: string2().min(1).describe("The project folder returned by prepare_footage."),
       layout: string2().default("full-frame"),
       caption: string2().default("two-tier"),
-      effects: array(object({ slug: string2(), start: number2().min(0), dur: number2().min(1.2).max(5).optional() }).passthrough()).max(12).default([]).describe("Same as render_reel: e.g. [{slug:'number-pop', start:3.2, number:'24', label:'hours later'}]."),
+      effects: array(object({ slug: string2(), start: number2().min(0), dur: number2().min(1.2).max(5).optional() }).passthrough()).max(24).default([]).describe("Same as render_reel: e.g. [{slug:'number-pop', start:3.2, number:'24', label:'hours later'}]."),
       images: array(string2()).max(8).optional().describe("Pictures for the layout, full paths on this computer."),
       emphasis: array(string2()).optional(),
       fixes: array(object({ from: string2(), to: string2() })).optional().describe("Transcript corrections; to:'' drops the word.")
@@ -22290,7 +22304,7 @@ ${r.next}` }, { type: "image", data: readFileSync4(r.storyboard).toString("base6
       project: string2().min(1).describe("The project folder returned by prepare_footage."),
       layout: string2().default("full-frame"),
       caption: string2().default("two-tier"),
-      effects: array(object({ slug: string2(), start: number2().min(0), dur: number2().min(1.2).max(5).optional() }).passthrough()).max(12).default([]).describe("e.g. [{slug:'number-pop', start:3.2, number:'24', label:'hours later'}]. Params per effect are in list_styles. Effect params image / images are picture files on this computer (full paths, jpg/png/webp); they are copied into the render folder, only the file name goes to Cutroom."),
+      effects: array(object({ slug: string2(), start: number2().min(0), dur: number2().min(1.2).max(5).optional() }).passthrough()).max(24).default([]).describe("e.g. [{slug:'number-pop', start:3.2, number:'24', label:'hours later'}]. Params per effect are in list_styles. Effect params image / images are picture files on this computer (full paths, jpg/png/webp); they are copied into the render folder, only the file name goes to Cutroom."),
       images: array(string2()).max(8).optional().describe("Pictures for the layout (image-cluster: 3 to 5, cutout-grid: 1 to 8, cutout-backdrop: optional): full paths of jpg, png or webp files on this computer. 8 pictures at most per reel, layout and effects together."),
       emphasis: array(string2()).optional().describe("Words to highlight in captions (otherwise numbers, then the longest word of each phrase)."),
       fixes: array(object({ from: string2(), to: string2() })).optional().describe("Transcript corrections; to:'' drops the word."),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@keepylabs/cutroom",
-  "version": "0.5.2",
+  "version": "0.5.4",
   "description": "Cutroom for Claude Code: hooks and scripts with scoring, reel breakdowns, creator teardowns, and finishing your own videos with QA. Needs a Keepy Labs membership key.",
   "type": "module",
   "bin": {