@ignix-social/editor 0.1.3 → 0.1.5

package/src/prepare.mjs

@@ -3,7 +3,7 @@
 //    con 4 decimales, 97/30 = 3.2333 daba el fotograma 96 en vez del 97);
 //  · ignix-frame.js de esta versión del editor, siempre al día.
 // El proyecto del usuario no se toca.
-import { copyFileSync, cpSync, existsSync, mkdirSync, readFileSync, readdirSync, realpathSync, rmSync, statSync, symlinkSync, writeFileSync } from 'node:fs';
+import { copyFileSync, cpSync, existsSync, linkSync, mkdirSync, readFileSync, readdirSync, realpathSync, rmSync, statSync, symlinkSync, writeFileSync } from 'node:fs';
 import { dirname, join } from 'node:path';
 import { fileURLToPath } from 'node:url';
 
@@ -12,6 +12,17 @@ const SKIP = new Set(['.ignix-build', 'renders', 'snapshots', 'node_modules', '.
 // Vídeo y audio no se copian (pueden pesar gigas): enlace a la ruta absoluta del original.
 const MEDIA = /\.(mp4|mov|m4v|webm|mkv|avi|wav|mp3|m4a|aac|ogg|flac)$/i;
 
+// macOS/Linux: enlace simbólico. Windows: los simbólicos piden permisos de administrador o el
+// modo desarrollador, así que enlace duro (mismo disco) y, si no se puede, una copia.
+export function linkMedia(target, to) {
+  if (process.platform !== 'win32') return symlinkSync(target, to);
+  try {
+    linkSync(target, to);
+  } catch {
+    copyFileSync(target, to);
+  }
+}
+
 function mirror(src, dest) {
   mkdirSync(dest, { recursive: true });
   for (const name of readdirSync(src)) {
@@ -20,7 +31,7 @@ function mirror(src, dest) {
     const to = join(dest, name);
     const st = statSync(from);
     if (st.isDirectory()) mirror(from, to);
-    else if (MEDIA.test(name)) symlinkSync(realpathSync(from), to);
+    else if (MEDIA.test(name)) linkMedia(realpathSync(from), to);
     else copyFileSync(from, to);
   }
 }

package/src/transcribe.mjs

@@ -3,7 +3,7 @@
 // el guion en las palabras y whisper en los tiempos: whisper confunde palabras («ADM», «tiene»)
 // y a veces se come el final de una frase.
 import { spawnSync } from 'node:child_process';
-import { mkdtempSync, readFileSync, rmSync } from 'node:fs';
+import { mkdtempSync, readFileSync, rmSync, statSync } from 'node:fs';
 import { tmpdir } from 'node:os';
 import { join } from 'node:path';
 
@@ -69,3 +69,67 @@ export function alignToScript(script, tokens, end) {
   }
   return out.map((w) => ({ t: w.t, s: +w.s.toFixed(3), e: +w.e.toFixed(3) }));
 }
+
+/**
+ * Transcripción completa para Creator: idioma detectado (si no se da), cada palabra con su
+ * confianza (media de sus tokens) y las frases (cortadas en un punto o en una pausa).
+ */
+export function transcribeDetailed({ ffmpeg, whisper, model, input, language = null, threads = 8, maxWords = 6000, maxSegments = 800 }) {
+  const tmp = mkdtempSync(join(tmpdir(), 'ignix-editor-asr-'));
+  try {
+    const wav = join(tmp, 'audio.wav');
+    const a = spawnSync(ffmpeg, ['-nostdin', '-hide_banner', '-loglevel', 'error', '-y', '-i', input, '-vn', '-ar', '16000', '-ac', '1', '-c:a', 'pcm_s16le', wav], { encoding: 'utf8' });
+    if (a.status !== 0) throw new Error(`Could not read the audio of ${input}: ${a.stderr.slice(-300)}`);
+    const duration = statSync(wav).size > 44 ? (statSync(wav).size - 44) / 32000 : 0;
+    const out = join(tmp, 'out');
+    const args = ['-m', model, '-l', language || 'auto', '-ml', '1', '-sow', '-ojf', '-of', out, '-t', String(threads), '-f', wav];
+    const run = (extra) => spawnSync(whisper, [...extra, ...args], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024 });
+    const intelMac = process.platform === 'darwin' && process.arch === 'x64';
+    let w = run(intelMac ? ['-ng'] : []);
+    let device = intelMac ? 'cpu' : 'gpu';
+    if (w.status !== 0 && !intelMac) {
+      w = run(['-ng']);
+      device = 'cpu';
+    }
+    if (w.status !== 0) throw new Error(`whisper failed: ${(w.stderr || '').slice(-400)}`);
+    const j = JSON.parse(readFileSync(`${out}.json`, 'utf8'));
+    const words = [];
+    let aside = false; // dentro de una acotación: [MÚSICA], (upbeat music)…
+    for (const seg of j.transcription ?? []) {
+      const text = seg.text.trim();
+      if (!text) continue;
+      if (aside || /^[[(]/.test(text)) {
+        aside = !/[\])]$/.test(text);
+        continue;
+      }
+      const toks = (seg.tokens ?? []).filter((t) => !/^\[_|^<\|/.test(t.text) && t.text.trim());
+      const p = toks.length ? toks.reduce((s, t) => s + (t.p ?? 0), 0) / toks.length : 0;
+      words.push({ text: text.slice(0, 60), start: seg.offsets.from / 1000, end: seg.offsets.to / 1000, p: +Math.min(1, Math.max(0, p)).toFixed(3) });
+      if (words.length >= maxWords) break;
+    }
+    // Frases: hasta un punto final o una pausa de más de 0,7 s.
+    const segments = [];
+    let cur = null;
+    for (const wd of words) {
+      if (cur && wd.start - cur.end > 0.7) {
+        segments.push(cur);
+        cur = null;
+      }
+      cur = cur ? { start: cur.start, end: wd.end, text: `${cur.text} ${wd.text}` } : { start: wd.start, end: wd.end, text: wd.text };
+      if (/[.!?…。]$/.test(wd.text)) {
+        segments.push(cur);
+        cur = null;
+      }
+    }
+    if (cur) segments.push(cur);
+    return {
+      language: j.result?.language ?? language ?? null,
+      duration: +duration.toFixed(3),
+      device,
+      words,
+      segments: segments.slice(0, maxSegments).map((s) => ({ ...s, text: s.text.slice(0, 500) })),
+    };
+  } finally {
+    rmSync(tmp, { recursive: true, force: true });
+  }
+}

package/vendor/hyperframes/UPSTREAM.json

@@ -7,6 +7,7 @@
   "patches": [
     "no-telemetry",
     "no-auto-update",
+    "mf-encoder",
     "no-desktop-promo"
   ]
 }

package/vendor/hyperframes/dist/chunk-K2FTXJYH.js

@@ -7405,14 +7405,16 @@ var GPU_ENCODER_CANDIDATES = [
   "videotoolbox",
   "vaapi",
   "qsv",
-  "amf"
+  "amf",
+  "mf" // ignix-editor: Media Foundation (Windows)
 ];
 var H264_ENCODER_BY_GPU = {
   nvenc: "h264_nvenc",
   videotoolbox: "h264_videotoolbox",
   vaapi: "h264_vaapi",
   qsv: "h264_qsv",
-  amf: "h264_amf"
+  amf: "h264_amf",
+  mf: "h264_mf"
 };
 var GPU_PROBE_TIMEOUT_MS = 2e3;
 var GPU_PROBE_KILL_GRACE_MS = 1e3;
@@ -7476,6 +7478,8 @@ function getGpuEncoderName(encoder, codec) {
       return codec === "h264" ? "h264_qsv" : "hevc_qsv";
     case "amf":
       return codec === "h264" ? "h264_amf" : "hevc_amf";
+    case "mf":
+      return codec === "h264" ? "h264_mf" : "hevc_mf";
     default:
       return codec === "h264" ? "libx264" : "libx265";
   }
@@ -8595,6 +8599,11 @@ function buildEncoderArgs(options, inputArgs, outputPath, gpuEncoder = null) {
           if (bitrate) args.push("-b:v", bitrate);
           else args.push("-rc", "cqp", "-qp_i", String(quality), "-qp_p", String(quality));
           break;
+        case "mf": {
+          const vt = buildVideoToolboxRateControlArgs({ bitrate, width: options.width, height: options.height, fps: fpsToNumber(fps), quality });
+          args.push("-b:v", vt[1], "-rate_control", "u_vbr");
+          break;
+        }
       }
       if (codec === "h264" && (gpuEncoder === "nvenc" || gpuEncoder === "qsv" || gpuEncoder === "vaapi" || gpuEncoder === "amf")) {
         args.push("-bf", "0");
@@ -9248,6 +9257,11 @@ function buildStreamingArgs(options, outputPath, gpuEncoder = null) {
           if (bitrate) args.push("-b:v", bitrate);
           else args.push("-rc", "cqp", "-qp_i", String(quality), "-qp_p", String(quality));
           break;
+        case "mf": {
+          const vt = buildVideoToolboxRateControlArgs({ bitrate, width: options.width, height: options.height, fps: fpsToNumber(fps), quality });
+          args.push("-b:v", vt[1], "-rate_control", "u_vbr");
+          break;
+        }
       }
       if (codec === "h264" && (gpuEncoder === "nvenc" || gpuEncoder === "qsv" || gpuEncoder === "vaapi" || gpuEncoder === "amf")) {
         args.push("-bf", "0");

package/vendor/hyperframes/dist/chunk-TVCRUG2S.js

@@ -195,7 +195,7 @@ var H264EncoderUnavailableError = class extends Error {
 function resolveH264EncoderMode(ffmpegEncodersOutput, gpuRequested) {
   if (gpuRequested) return "gpu";
   if (/\blibx264\b/.test(ffmpegEncodersOutput)) return "software";
-  if (/\bh264_videotoolbox\b/.test(ffmpegEncodersOutput)) return "gpu";
+  if (/\bh264_(videotoolbox|mf)\b/.test(ffmpegEncodersOutput)) return "gpu";
   throw new H264EncoderUnavailableError();
 }
 function detectH264EncoderMode(ffmpegPath, gpuRequested) {

package/vendor/hyperframes/dist/render-LNO62LEL.js

@@ -1770,7 +1770,7 @@ async function executeLocalRender(projectDir, outputPath, options, cancellation)
     }
     if (encoderMode === "gpu") {
       console.warn(
-        c.warn("  FFmpeg does not include libx264; falling back to VideoToolbox H.264 encoding.")
+        c.warn("  FFmpeg does not include libx264; using the system's hardware H.264 encoder (VideoToolbox on macOS, Media Foundation on Windows).")
       );
       options = { ...options, gpu: true };
     }

package/src/native/faces.swift

@@ -1,60 +0,0 @@
-// faces — dónde está la cara en cada momento de un vídeo (Apple Vision, macOS 13+).
-// Uso: faces <vídeo> <cada_segundos>  → JSON por stdout:
-//   { "duration": s, "width": px, "height": px, "samples": [ { "t": s, "x":…, "y":…, "w":…, "h":… } | { "t": s } ] }
-// x, y, w, h normalizados 0–1 con origen ARRIBA a la izquierda (como CSS). Sin cara → solo "t".
-import AVFoundation
-import Foundation
-import Vision
-
-@main
-struct Faces {
-  static func main() async {
-    let args = CommandLine.arguments
-    guard args.count >= 2 else {
-      FileHandle.standardError.write("usage: faces <video> [every_seconds]\n".data(using: .utf8)!)
-      exit(2)
-    }
-    let url = URL(fileURLWithPath: args[1])
-    let every = args.count >= 3 ? (Double(args[2]) ?? 0.2) : 0.2
-    let asset = AVURLAsset(url: url)
-    do {
-      let duration = try await asset.load(.duration).seconds
-      guard let track = try await asset.loadTracks(withMediaType: .video).first else {
-        FileHandle.standardError.write("no video track\n".data(using: .utf8)!)
-        exit(1)
-      }
-      let size = try await track.load(.naturalSize).applying(try await track.load(.preferredTransform))
-      let gen = AVAssetImageGenerator(asset: asset)
-      gen.appliesPreferredTrackTransform = true
-      gen.requestedTimeToleranceBefore = CMTime(value: 1, timescale: 30)
-      gen.requestedTimeToleranceAfter = CMTime(value: 1, timescale: 30)
-      gen.maximumSize = CGSize(width: 960, height: 960)
-
-      var samples: [[String: Double]] = []
-      var t = 0.0
-      while t < duration {
-        var sample: [String: Double] = ["t": (t * 1000).rounded() / 1000]
-        if let (image, _) = try? await gen.image(at: CMTime(seconds: t, preferredTimescale: 600)) {
-          let req = VNDetectFaceRectanglesRequest()
-          try? VNImageRequestHandler(cgImage: image, options: [:]).perform([req])
-          // La cara más grande: la de quien habla a cámara.
-          if let f = (req.results ?? []).max(by: { $0.boundingBox.width * $0.boundingBox.height < $1.boundingBox.width * $1.boundingBox.height }) {
-            let b = f.boundingBox // Vision: origen ABAJO a la izquierda
-            sample["x"] = b.minX
-            sample["y"] = 1 - b.maxY
-            sample["w"] = b.width
-            sample["h"] = b.height
-          }
-        }
-        samples.append(sample)
-        t += every
-      }
-      let out: [String: Any] = ["duration": duration, "width": abs(size.width), "height": abs(size.height), "samples": samples]
-      let data = try JSONSerialization.data(withJSONObject: out)
-      FileHandle.standardOutput.write(data)
-    } catch {
-      FileHandle.standardError.write("faces failed: \(error)\n".data(using: .utf8)!)
-      exit(1)
-    }
-  }
-}