|
| 1 | +import { test } from "node:test"; |
| 2 | +import assert from "node:assert/strict"; |
| 3 | +import { spawnSync } from "node:child_process"; |
| 4 | +import { mkdtempSync, readFileSync, writeFileSync, rmSync } from "node:fs"; |
| 5 | +import { join, dirname } from "node:path"; |
| 6 | +import { tmpdir } from "node:os"; |
| 7 | +import { fileURLToPath } from "node:url"; |
| 8 | +import { atempoChain, parseWords, scaleWords } from "./lib/recording.mjs"; |
| 9 | + |
| 10 | +const AUDIO = join(dirname(fileURLToPath(import.meta.url)), "audio.mjs"); |
| 11 | + |
| 12 | +test("atempoChain splits speeds outside ffmpeg's 0.5..2 stage range", () => { |
| 13 | + assert.equal(atempoChain(1.5), "atempo=1.5"); |
| 14 | + assert.equal(atempoChain(3), "atempo=2,atempo=1.5"); |
| 15 | + assert.equal(atempoChain(0.25), "atempo=0.5,atempo=0.5"); |
| 16 | +}); |
| 17 | + |
| 18 | +test("parseWords takes text or word and rejects a missing time", () => { |
| 19 | + assert.deepEqual(parseWords([{ word: "hi", start: 0, end: 1 }]), [ |
| 20 | + { text: "hi", start: 0, end: 1 }, |
| 21 | + ]); |
| 22 | + assert.equal(parseWords(undefined), null); |
| 23 | + assert.throws(() => parseWords([{ text: "hi", start: 0 }]), /words\[0\]/); |
| 24 | +}); |
| 25 | + |
| 26 | +test("scaleWords moves words by 1/speed", () => { |
| 27 | + assert.deepEqual(scaleWords([{ text: "a", start: 1, end: 2 }], 2), [ |
| 28 | + { text: "a", start: 0.5, end: 1 }, |
| 29 | + ]); |
| 30 | +}); |
| 31 | + |
| 32 | +test("a recording line with malformed words is omitted with the reason, not a crash", () => { |
| 33 | + const dir = mkdtempSync(join(tmpdir(), "hf-rec-")); |
| 34 | + try { |
| 35 | + const make = spawnSync( |
| 36 | + "ffmpeg", |
| 37 | + ["-y", "-loglevel", "error", "-f", "lavfi", "-i", "sine=duration=1", join(dir, "take.wav")], |
| 38 | + { encoding: "utf8" }, |
| 39 | + ); |
| 40 | + assert.equal(make.status, 0, make.stderr); |
| 41 | + writeFileSync( |
| 42 | + join(dir, "audio_request.json"), |
| 43 | + JSON.stringify({ |
| 44 | + lines: [{ id: "01", audio: "take.wav", words: [{ text: "x", start: 0 }] }], |
| 45 | + }), |
| 46 | + ); |
| 47 | + const run = spawnSync( |
| 48 | + process.execPath, |
| 49 | + [ |
| 50 | + AUDIO, |
| 51 | + "--request", |
| 52 | + join(dir, "audio_request.json"), |
| 53 | + "--hyperframes", |
| 54 | + dir, |
| 55 | + "--out", |
| 56 | + join(dir, "m.json"), |
| 57 | + "--only", |
| 58 | + "tts", |
| 59 | + ], |
| 60 | + { encoding: "utf8", env: { ...process.env, HOME: dir } }, |
| 61 | + ); |
| 62 | + assert.equal(run.status, 0, run.stderr); |
| 63 | + assert.match(run.stdout, /line 01: recording failed — omitted \(words\[0\] needs text/); |
| 64 | + assert.deepEqual(JSON.parse(readFileSync(join(dir, "m.json"), "utf8")).voices, []); |
| 65 | + } finally { |
| 66 | + rmSync(dir, { recursive: true, force: true }); |
| 67 | + } |
| 68 | +}); |
| 69 | + |
| 70 | +// A 1 s gap then a 1 s tone is a "word" at 1..2 s. At speed 2 the tone must start at 0.5 s in the |
| 71 | +// stretched file, which is where the scaled word says it is. |
| 72 | +test("a recording at speed 2: the file is stretched and the words still land on the audio", () => { |
| 73 | + const dir = mkdtempSync(join(tmpdir(), "hf-rec-")); |
| 74 | + try { |
| 75 | + const make = spawnSync( |
| 76 | + "ffmpeg", |
| 77 | + [ |
| 78 | + "-y", |
| 79 | + "-loglevel", |
| 80 | + "error", |
| 81 | + "-f", |
| 82 | + "lavfi", |
| 83 | + "-i", |
| 84 | + "anullsrc=r=44100:cl=mono", |
| 85 | + "-f", |
| 86 | + "lavfi", |
| 87 | + "-i", |
| 88 | + "sine=frequency=440:duration=1:sample_rate=44100", |
| 89 | + "-filter_complex", |
| 90 | + "[0]atrim=duration=1[s];[s][1]concat=n=2:v=0:a=1", |
| 91 | + join(dir, "take.wav"), |
| 92 | + ], |
| 93 | + { encoding: "utf8" }, |
| 94 | + ); |
| 95 | + assert.equal(make.status, 0, make.stderr); |
| 96 | + writeFileSync( |
| 97 | + join(dir, "audio_request.json"), |
| 98 | + JSON.stringify({ |
| 99 | + speed: 2, |
| 100 | + lines: [{ id: "01", audio: "take.wav", words: [{ text: "tone", start: 1, end: 2 }] }], |
| 101 | + }), |
| 102 | + ); |
| 103 | + const run = spawnSync( |
| 104 | + process.execPath, |
| 105 | + [ |
| 106 | + AUDIO, |
| 107 | + "--request", |
| 108 | + join(dir, "audio_request.json"), |
| 109 | + "--hyperframes", |
| 110 | + dir, |
| 111 | + "--out", |
| 112 | + join(dir, "audio_meta.json"), |
| 113 | + "--only", |
| 114 | + "tts", |
| 115 | + ], |
| 116 | + { encoding: "utf8", env: { ...process.env, HOME: dir } }, |
| 117 | + ); |
| 118 | + assert.equal(run.status, 0, run.stderr + run.stdout); |
| 119 | + const meta = JSON.parse(readFileSync(join(dir, "audio_meta.json"), "utf8")); |
| 120 | + const voice = meta.voices[0]; |
| 121 | + assert.equal(meta.tts_provider, "recording"); |
| 122 | + assert.equal(voice.source, "recording"); |
| 123 | + assert.ok(Math.abs(voice.duration_s - 1) < 0.05, `duration ${voice.duration_s}`); |
| 124 | + assert.deepEqual([voice.words[0].start, voice.words[0].end], [0.5, 1]); |
| 125 | + const probe = spawnSync( |
| 126 | + "ffmpeg", |
| 127 | + ["-i", join(dir, voice.path), "-af", "silencedetect=noise=-40dB:d=0.2", "-f", "null", "-"], |
| 128 | + { encoding: "utf8" }, |
| 129 | + ); |
| 130 | + const silenceEnd = Number(/silence_end: ([\d.]+)/.exec(probe.stderr)?.[1]); |
| 131 | + assert.ok( |
| 132 | + Math.abs(silenceEnd - voice.words[0].start) < 0.08, |
| 133 | + `tone starts at ${silenceEnd}s, word says ${voice.words[0].start}s`, |
| 134 | + ); |
| 135 | + } finally { |
| 136 | + rmSync(dir, { recursive: true, force: true }); |
| 137 | + } |
| 138 | +}); |
0 commit comments