@ignix-social/editor 0.1.17 → 0.1.19
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/package.json +1 -1
- package/skills/ignix-editor/SKILL.md +16 -0
- package/src/cli.mjs +31 -0
- package/src/giphy.mjs +136 -0
- package/src/transcribe.mjs +29 -7
package/package.json
CHANGED
|
@@ -16,6 +16,7 @@ Commands. `ignix-editor` means the copy the Ignix playbook installed: `~/.ignix-
|
|
|
16
16
|
| `ignix-editor transcribe <file> --lang es [--script "…"] -o words.json` | Words with start/end seconds, on this machine. |
|
|
17
17
|
| `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. |
|
|
18
18
|
| `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. |
|
|
19
|
+
| `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). |
|
|
19
20
|
| `ignix-editor check [dir]` | HTML well-formed, no network assets, media rules, layout/contrast/motion audit. |
|
|
20
21
|
| `ignix-editor snapshot [dir] --at 0,1.5,3` | PNG frames into `<dir>/snapshots/` to LOOK at. `[dir]` defaults to the current folder. |
|
|
21
22
|
| `ignix-editor render [dir] [-o out.mp4]` | Renders to `<dir>/renders/<dir>.mp4`, audio normalised to −14 LUFS. |
|
|
@@ -76,6 +77,21 @@ The renderer asks for frames independently (in parallel, out of order). If frame
|
|
|
76
77
|
- Use `transform` (translate/scale/rotate) and `opacity` for motion; animating `left/top/width/height/font-size` snaps to whole pixels and stutters in slow moves.
|
|
77
78
|
- Toggling `display`, `textContent` or a class from `t` is fine (typed text: `KEY.slice(0, n(t))`).
|
|
78
79
|
|
|
80
|
+
### GIPHY stickers (transparent, animated)
|
|
81
|
+
|
|
82
|
+
Never put an animated GIF/WebP in an `<img>`: it plays on the browser's clock, not the render's, and comes out jumpy. `ignix-editor giphy <sticker link>` gives a transparent sheet with every frame and prints `frames`, `cols`, `rows`, `fps` and the frame size. Show one cell at a time, chosen from `t`:
|
|
83
|
+
|
|
84
|
+
```js
|
|
85
|
+
// e.g. 51 frames, 8×7 sheet, 25 fps, 256×256 each; the element is 256×256 (scale it with transform)
|
|
86
|
+
const S = { frames: 51, cols: 8, rows: 7, fps: 25 };
|
|
87
|
+
sticker.style.background = `url(assets/giphy-sticker-ID-sheet.png) 0 0 / ${S.cols * 100}% ${S.rows * 100}% no-repeat`;
|
|
88
|
+
// inside run(): loops while it is on screen
|
|
89
|
+
const i = Math.floor(Math.max(0, t - start) * S.fps) % S.frames;
|
|
90
|
+
sticker.style.backgroundPosition = `${((i % S.cols) / (S.cols - 1)) * 100}% ${(Math.floor(i / S.cols) / (S.rows - 1)) * 100}%`;
|
|
91
|
+
```
|
|
92
|
+
|
|
93
|
+
GIFs and clips come as MP4: use them like any `<video>` (a clip keeps its audio; mute it if the voice-over runs over it).
|
|
94
|
+
|
|
79
95
|
### Never measure the page while rendering
|
|
80
96
|
|
|
81
97
|
Do not call `getBoundingClientRect()`/`offsetLeft` to aim arrows, cursors or highlight boxes: fonts may not be ready and layout may differ per frame; arrows pointed 300 px off and vanished in a test because of this. Instead, give the recreated UI fixed coordinates and keep a table:
|
package/src/cli.mjs
CHANGED
|
@@ -20,6 +20,7 @@ import { fetchAll } from './fetch.mjs';
|
|
|
20
20
|
import { measureStyle } from './analysis/style.mjs';
|
|
21
21
|
import { reviewVideo } from './analysis/review.mjs';
|
|
22
22
|
import { pickAudio } from './pick-audio.mjs';
|
|
23
|
+
import { importGiphy, openGiphySearch } from './giphy.mjs';
|
|
23
24
|
|
|
24
25
|
const PKG_DIR = join(dirname(fileURLToPath(import.meta.url)), '..');
|
|
25
26
|
const PKG = JSON.parse(readFileSync(join(PKG_DIR, 'package.json'), 'utf8'));
|
|
@@ -251,6 +252,34 @@ const commands = {
|
|
|
251
252
|
}
|
|
252
253
|
},
|
|
253
254
|
|
|
255
|
+
// GIFs, stickers y clips de GIPHY elegidos por la persona (src/giphy.mjs): abre la
|
|
256
|
+
// búsqueda en su navegador, o trae el enlace que ha copiado.
|
|
257
|
+
async giphy(argv) {
|
|
258
|
+
const { pos, opts } = parseArgs(argv, { search: 'str', kind: 'str', open: 'bool', json: 'bool' });
|
|
259
|
+
const usage = 'Usage: ignix-editor giphy --search "thumbs up" [--kind gifs|stickers|clips] or ignix-editor giphy <link> [folder]';
|
|
260
|
+
if (opts.search) {
|
|
261
|
+
const kind = ['stickers', 'clips'].includes(opts.kind) ? opts.kind : 'gifs';
|
|
262
|
+
const r = opts.open === false ? { url: (await import('./giphy.mjs')).giphySearchUrl(opts.search, kind), opened: false } : openGiphySearch(opts.search, kind);
|
|
263
|
+
log(r.opened ? `Opened in the browser: ${r.url}` : `Open this in the browser: ${r.url}`);
|
|
264
|
+
log('Ask the person to copy the link of the one they like (Share → Copy link) and paste it here; then run: ignix-editor giphy <link> <folder>');
|
|
265
|
+
if (opts.json) console.log(JSON.stringify(r));
|
|
266
|
+
return;
|
|
267
|
+
}
|
|
268
|
+
const link = pos[0];
|
|
269
|
+
if (!link) fail(usage);
|
|
270
|
+
const project = resolve(pos[1] ?? '.');
|
|
271
|
+
if (!existsSync(project)) fail(`Folder not found: ${project}`);
|
|
272
|
+
const forced = { gifs: 'gif', gif: 'gif', stickers: 'sticker', sticker: 'sticker', clips: 'clip', clip: 'clip' }[opts.kind];
|
|
273
|
+
const ffmpeg = await ensureTool('ffmpeg', log);
|
|
274
|
+
const ffprobe = await ensureTool('ffprobe', log);
|
|
275
|
+
const r = await importGiphy({ link, project, kind: forced, ffmpeg, ffprobe }).catch((e) => fail(e.message));
|
|
276
|
+
if (r.sheet) {
|
|
277
|
+
const sh = r.sheet;
|
|
278
|
+
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).`);
|
|
279
|
+
} else log(`✓ ${r.kind} → ${r.file}${r.hasAudio ? ' (with its audio)' : ''}`);
|
|
280
|
+
if (opts.json) console.log(JSON.stringify(r));
|
|
281
|
+
},
|
|
282
|
+
|
|
254
283
|
// El estilo de edición de alguien, medido en 2–3 vídeos suyos (ADR-0023 E3).
|
|
255
284
|
async style(argv) {
|
|
256
285
|
const [sub, ...rest] = argv;
|
|
@@ -390,6 +419,8 @@ const commands = {
|
|
|
390
419
|
ignix-editor faces <video> [--every 0.2] [-o faces.json] where the face is, smoothed
|
|
391
420
|
ignix-editor pick-audio [folder] --kind music|sfx --query "inspiring" [--order trending]
|
|
392
421
|
opens the Pixabay search in the browser, waits for the person's download, copies it into assets/
|
|
422
|
+
ignix-editor giphy --search "thumbs up" [--kind gifs|stickers|clips] opens the GIPHY search in the browser
|
|
423
|
+
ignix-editor giphy <link> [folder] brings the GIF (mp4), sticker (transparent frame sheet) or clip (mp4 with audio) into assets/
|
|
393
424
|
ignix-editor style measure <video>… [--speech] [--stills dir] [-o style.json] someone's editing style, measured on 2–3 of their videos
|
|
394
425
|
ignix-editor ffmpeg|ffprobe <args> Ignix Editor's own ffmpeg
|
|
395
426
|
ignix-editor skills add installs the skill for Claude
|
package/src/giphy.mjs
ADDED
|
@@ -0,0 +1,136 @@
|
|
|
1
|
+
import { spawnSync } from 'node:child_process';
|
|
2
|
+
import { createWriteStream, existsSync, mkdirSync, readFileSync, rmSync, writeFileSync } from 'node:fs';
|
|
3
|
+
import { join } from 'node:path';
|
|
4
|
+
import { Readable } from 'node:stream';
|
|
5
|
+
import { pipeline } from 'node:stream/promises';
|
|
6
|
+
import { openInBrowser } from './pick-audio.mjs';
|
|
7
|
+
|
|
8
|
+
// GIFs, STICKERS Y CLIPS DE GIPHY, ELEGIDOS POR LA PERSONA (2026-10-08, Fran).
|
|
9
|
+
//
|
|
10
|
+
// Sin la API de GIPHY (pide clave y revisión): la persona busca en giphy.com (esto le
|
|
11
|
+
// abre la búsqueda), copia el enlace del que le gusta y el agente lo trae con
|
|
12
|
+
// `ignix-editor giphy <enlace>`. Con el id del enlace, GIPHY sirve cada formato en
|
|
13
|
+
// media.giphy.com: el GIF como MP4 (mejor para vídeo), el sticker como WebP con
|
|
14
|
+
// transparencia (para superponerlo) y el clip como MP4 con su audio.
|
|
15
|
+
//
|
|
16
|
+
// El sticker NO va como imagen animada: un GIF/WebP en un <img> corre con el reloj del
|
|
17
|
+
// navegador, no con el del render (cada fotograma se pide suelto y fuera de orden), así
|
|
18
|
+
// que saldría a saltos. Va como TIRA de fotogramas en un PNG con transparencia
|
|
19
|
+
// (ffmpeg, filtro tile) que el vídeo recorre desde `t` (ver la guía, sección GIPHY).
|
|
20
|
+
|
|
21
|
+
const ID = /^[A-Za-z0-9]{6,40}$/;
|
|
22
|
+
|
|
23
|
+
/** El id y el tipo de un enlace de GIPHY (página, media, i.giphy o embed). null si no lo es. */
|
|
24
|
+
export function parseGiphyLink(raw) {
|
|
25
|
+
let u;
|
|
26
|
+
try {
|
|
27
|
+
u = new URL(String(raw ?? '').trim());
|
|
28
|
+
} catch {
|
|
29
|
+
return null;
|
|
30
|
+
}
|
|
31
|
+
const host = u.hostname.replace(/^www\./, '');
|
|
32
|
+
if (!/(^|\.)giphy\.com$/.test(host)) return null;
|
|
33
|
+
const parts = u.pathname.split('/').filter(Boolean);
|
|
34
|
+
// giphy.com/gifs/<slug>-<id>, /stickers/<slug>-<id>, /clips/<slug>-<id>, /embed/<id>
|
|
35
|
+
if (host === 'giphy.com' && parts.length >= 2) {
|
|
36
|
+
const kind = { gifs: 'gif', stickers: 'sticker', clips: 'clip', embed: 'gif' }[parts[0]];
|
|
37
|
+
const last = parts[1];
|
|
38
|
+
const id = last.includes('-') ? last.slice(last.lastIndexOf('-') + 1) : last;
|
|
39
|
+
return kind && ID.test(id) ? { id, kind } : null;
|
|
40
|
+
}
|
|
41
|
+
// media*.giphy.com/media/(v1.xxx/)?<id>/<fichero>
|
|
42
|
+
const m = parts.indexOf('media');
|
|
43
|
+
if (m >= 0) {
|
|
44
|
+
const rest = parts.slice(m + 1);
|
|
45
|
+
const id = rest[0]?.startsWith('v1.') ? rest[1] : rest[0];
|
|
46
|
+
return id && ID.test(id) ? { id, kind: null } : null;
|
|
47
|
+
}
|
|
48
|
+
// i.giphy.com/<id>.webp
|
|
49
|
+
if (host === 'i.giphy.com' && parts[0]) {
|
|
50
|
+
const id = parts[0].replace(/\.[a-z0-9]+$/i, '');
|
|
51
|
+
return ID.test(id) ? { id, kind: null } : null;
|
|
52
|
+
}
|
|
53
|
+
return null;
|
|
54
|
+
}
|
|
55
|
+
|
|
56
|
+
/** La búsqueda de GIPHY como la usa una persona en su web. */
|
|
57
|
+
export function giphySearchUrl(query, kind = 'gifs') {
|
|
58
|
+
const slug = String(query ?? '')
|
|
59
|
+
.trim()
|
|
60
|
+
.toLowerCase()
|
|
61
|
+
.replace(/[^\p{L}\p{N}]+/gu, '-')
|
|
62
|
+
.replace(/^-|-$/g, '')
|
|
63
|
+
.slice(0, 80);
|
|
64
|
+
const suffix = kind === 'stickers' ? '-stickers' : kind === 'clips' ? '-clips' : '';
|
|
65
|
+
return `https://giphy.com/search/${encodeURIComponent(slug).replace(/%2D/gi, '-')}${suffix}`;
|
|
66
|
+
}
|
|
67
|
+
|
|
68
|
+
/** Qué fichero traer de media.giphy.com para cada tipo (y el de reserva). */
|
|
69
|
+
export function giphyRenditions(kind) {
|
|
70
|
+
if (kind === 'clip') return [['giphy480p.mp4', 'mp4'], ['giphy.mp4', 'mp4']];
|
|
71
|
+
if (kind === 'sticker') return [['giphy.gif', 'gif']];
|
|
72
|
+
return [['giphy.mp4', 'mp4'], ['giphy.gif', 'gif']];
|
|
73
|
+
}
|
|
74
|
+
|
|
75
|
+
/** Trae el elegido a `<project>/assets/` y apunta su origen en `work/media-sources.json`. */
|
|
76
|
+
/**
|
|
77
|
+
* Un GIF con transparencia → una tira PNG (cols × rows fotogramas) + sus medidas.
|
|
78
|
+
* `ffmpeg`/`ffprobe` son los del Editor.
|
|
79
|
+
*/
|
|
80
|
+
export function gifToSheet({ ffmpeg, ffprobe, gif, out }) {
|
|
81
|
+
const pr = spawnSync(ffprobe, ['-v', 'error', '-count_frames', '-select_streams', 'v:0', '-show_entries', 'stream=width,height,nb_read_frames,r_frame_rate:format=duration', '-of', 'json', gif], { encoding: 'utf8' });
|
|
82
|
+
const j = JSON.parse(pr.stdout || '{}');
|
|
83
|
+
const st = j.streams?.[0] ?? {};
|
|
84
|
+
const frames = Math.max(1, Number(st.nb_read_frames) || 1);
|
|
85
|
+
const durationSec = Number(j.format?.duration) || frames / 15;
|
|
86
|
+
const fps = Math.round((frames / durationSec) * 100) / 100;
|
|
87
|
+
// Rejilla casi cuadrada: una tira larga pasaría del tamaño máximo de textura de Chrome.
|
|
88
|
+
const cols = Math.ceil(Math.sqrt(frames));
|
|
89
|
+
const rows = Math.ceil(frames / cols);
|
|
90
|
+
const r = spawnSync(ffmpeg, ['-nostdin', '-y', '-v', 'error', '-i', gif, '-vf', `format=rgba,tile=${cols}x${rows}`, '-frames:v', '1', out], { encoding: 'utf8' });
|
|
91
|
+
if (r.status !== 0) throw new Error(`Could not turn the sticker into frames: ${(r.stderr || '').slice(-300)}`);
|
|
92
|
+
return { sheet: out, frames, cols, rows, fps, durationSec: Math.round(durationSec * 100) / 100, frameWidth: Number(st.width), frameHeight: Number(st.height) };
|
|
93
|
+
}
|
|
94
|
+
|
|
95
|
+
export async function importGiphy({ link, project, kind: forced, ffmpeg, ffprobe }) {
|
|
96
|
+
const parsed = parseGiphyLink(link);
|
|
97
|
+
if (!parsed) throw new Error('That is not a GIPHY link. Copy the link of the GIF, sticker or clip from giphy.com.');
|
|
98
|
+
const kind = forced ?? parsed.kind ?? 'gif';
|
|
99
|
+
const assets = join(project, 'assets');
|
|
100
|
+
mkdirSync(assets, { recursive: true });
|
|
101
|
+
let lastErr = null;
|
|
102
|
+
for (const [file, ext] of giphyRenditions(kind)) {
|
|
103
|
+
const url = `https://media.giphy.com/media/${parsed.id}/${file}`;
|
|
104
|
+
const res = await fetch(url).catch((e) => {
|
|
105
|
+
lastErr = e;
|
|
106
|
+
return null;
|
|
107
|
+
});
|
|
108
|
+
if (!res?.ok || !res.body) continue;
|
|
109
|
+
let dest = join(assets, `giphy-${kind}-${parsed.id}.${ext}`);
|
|
110
|
+
await pipeline(Readable.fromWeb(res.body), createWriteStream(dest));
|
|
111
|
+
let sheet = null;
|
|
112
|
+
if (kind === 'sticker' && ffmpeg && ffprobe) {
|
|
113
|
+
sheet = gifToSheet({ ffmpeg, ffprobe, gif: dest, out: join(assets, `giphy-sticker-${parsed.id}-sheet.png`) });
|
|
114
|
+
rmSync(dest, { force: true });
|
|
115
|
+
dest = sheet.sheet;
|
|
116
|
+
}
|
|
117
|
+
const work = join(project, 'work');
|
|
118
|
+
mkdirSync(work, { recursive: true });
|
|
119
|
+
const log = join(work, 'media-sources.json');
|
|
120
|
+
let prev = [];
|
|
121
|
+
try {
|
|
122
|
+
prev = existsSync(log) ? JSON.parse(readFileSync(log, 'utf8')) : [];
|
|
123
|
+
} catch {
|
|
124
|
+
/* fichero roto: se empieza de nuevo */
|
|
125
|
+
}
|
|
126
|
+
writeFileSync(log, JSON.stringify([...prev, { source: 'giphy', kind, id: parsed.id, link: String(link), file: dest, at: new Date().toISOString() }], null, 2));
|
|
127
|
+
return { id: parsed.id, kind, file: dest, hasAudio: kind === 'clip' && file.includes('480p'), ...(sheet ? { sheet } : {}) };
|
|
128
|
+
}
|
|
129
|
+
throw new Error(`Could not download that ${kind} from GIPHY${lastErr ? `: ${lastErr.message}` : ''}.`);
|
|
130
|
+
}
|
|
131
|
+
|
|
132
|
+
/** Abre la búsqueda en el navegador de la persona. */
|
|
133
|
+
export function openGiphySearch(query, kind) {
|
|
134
|
+
const url = giphySearchUrl(query, kind);
|
|
135
|
+
return { url, opened: openInBrowser(url) };
|
|
136
|
+
}
|
package/src/transcribe.mjs
CHANGED
|
@@ -5,19 +5,40 @@
|
|
|
5
5
|
import { spawnSync } from 'node:child_process';
|
|
6
6
|
import { mkdtempSync, readFileSync, rmSync, statSync } from 'node:fs';
|
|
7
7
|
import { tmpdir } from 'node:os';
|
|
8
|
-
import { join } from 'node:path';
|
|
8
|
+
import { basename, dirname, join, relative } from 'node:path';
|
|
9
|
+
|
|
10
|
+
/**
|
|
11
|
+
* Dónde trabaja whisper. En Windows, whisper.cpp NO abre rutas con caracteres fuera del ASCII
|
|
12
|
+
* (pasó con «C:\Users\José María»: el modelo, el audio y la salida cuelgan todos de esa
|
|
13
|
+
* carpeta, %TEMP% incluido). Así que se le ejecuta DESDE la carpeta del modelo y se le pasan
|
|
14
|
+
* solo rutas relativas, que son ASCII: el modelo por su nombre y el audio y la salida en una
|
|
15
|
+
* subcarpeta temporal de ahí. ffmpeg y el arranque del proceso sí entienden Unicode.
|
|
16
|
+
*/
|
|
17
|
+
function whisperWorkspace(model) {
|
|
18
|
+
if (process.platform === 'win32') {
|
|
19
|
+
try {
|
|
20
|
+
const cwd = dirname(model);
|
|
21
|
+
const tmp = mkdtempSync(join(cwd, 'asr-'));
|
|
22
|
+
return { tmp, cwd, modelArg: basename(model), arg: (p) => relative(cwd, p) };
|
|
23
|
+
} catch {
|
|
24
|
+
/* carpeta del modelo sin permiso de escritura: como en el resto de sistemas */
|
|
25
|
+
}
|
|
26
|
+
}
|
|
27
|
+
return { tmp: mkdtempSync(join(tmpdir(), 'ignix-editor-asr-')), cwd: undefined, modelArg: model, arg: (p) => p };
|
|
28
|
+
}
|
|
9
29
|
|
|
10
30
|
const norm = (s) => s.toLowerCase().normalize('NFD').replace(/[̀-ͯ]/g, '').replace(/[^\p{L}\p{N}]/gu, '');
|
|
11
31
|
|
|
12
32
|
export function transcribe({ ffmpeg, whisper, model, input, language = 'es', threads = 8 }) {
|
|
13
|
-
const
|
|
33
|
+
const ws = whisperWorkspace(model);
|
|
34
|
+
const tmp = ws.tmp;
|
|
14
35
|
try {
|
|
15
36
|
const wav = join(tmp, 'audio.wav');
|
|
16
37
|
const a = spawnSync(ffmpeg, ['-nostdin', '-hide_banner', '-loglevel', 'error', '-y', '-i', input, '-vn', '-ar', '16000', '-ac', '1', '-c:a', 'pcm_s16le', wav], { encoding: 'utf8' });
|
|
17
38
|
if (a.status !== 0) throw new Error(`Could not read the audio of ${input}: ${a.stderr.slice(-300)}`);
|
|
18
39
|
const out = join(tmp, 'out');
|
|
19
|
-
const args = ['-m',
|
|
20
|
-
const run = (extra) => spawnSync(whisper, [...extra, ...args], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024 });
|
|
40
|
+
const args = ['-m', ws.modelArg, '-l', language, '-ml', '1', '-sow', '-oj', '-of', ws.arg(out), '-t', String(threads), '-f', ws.arg(wav)];
|
|
41
|
+
const run = (extra) => spawnSync(whisper, [...extra, ...args], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024, cwd: ws.cwd });
|
|
21
42
|
// En Mac Intel la GPU (Metal) se cuelga con whisper: CPU directamente. En el resto, GPU y,
|
|
22
43
|
// si falla, otra vez por CPU.
|
|
23
44
|
const intelMac = process.platform === 'darwin' && process.arch === 'x64';
|
|
@@ -75,15 +96,16 @@ export function alignToScript(script, tokens, end) {
|
|
|
75
96
|
* confianza (media de sus tokens) y las frases (cortadas en un punto o en una pausa).
|
|
76
97
|
*/
|
|
77
98
|
export function transcribeDetailed({ ffmpeg, whisper, model, input, language = null, threads = 8, maxWords = 6000, maxSegments = 800 }) {
|
|
78
|
-
const
|
|
99
|
+
const ws = whisperWorkspace(model);
|
|
100
|
+
const tmp = ws.tmp;
|
|
79
101
|
try {
|
|
80
102
|
const wav = join(tmp, 'audio.wav');
|
|
81
103
|
const a = spawnSync(ffmpeg, ['-nostdin', '-hide_banner', '-loglevel', 'error', '-y', '-i', input, '-vn', '-ar', '16000', '-ac', '1', '-c:a', 'pcm_s16le', wav], { encoding: 'utf8' });
|
|
82
104
|
if (a.status !== 0) throw new Error(`Could not read the audio of ${input}: ${a.stderr.slice(-300)}`);
|
|
83
105
|
const duration = statSync(wav).size > 44 ? (statSync(wav).size - 44) / 32000 : 0;
|
|
84
106
|
const out = join(tmp, 'out');
|
|
85
|
-
const args = ['-m',
|
|
86
|
-
const run = (extra) => spawnSync(whisper, [...extra, ...args], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024 });
|
|
107
|
+
const args = ['-m', ws.modelArg, '-l', language || 'auto', '-ml', '1', '-sow', '-ojf', '-of', ws.arg(out), '-t', String(threads), '-f', ws.arg(wav)];
|
|
108
|
+
const run = (extra) => spawnSync(whisper, [...extra, ...args], { encoding: 'utf8', maxBuffer: 64 * 1024 * 1024, cwd: ws.cwd });
|
|
87
109
|
const intelMac = process.platform === 'darwin' && process.arch === 'x64';
|
|
88
110
|
let w = run(intelMac ? ['-ng'] : []);
|
|
89
111
|
let device = intelMac ? 'cpu' : 'gpu';
|