Clone any viral video with AI agents. Not just a script, the whole workflow: swap the face, the words, the B-roll, ship 100 variants in one command, and get your 100M views.
TypeScript agentic-aiaiai-agentsai-videocompilerdslffmpeggenerative-aillmmarkup-languagemonorepoplugin-systemprogramming-languagetext-to-videotypescriptvideovideo-automationvideo-clonevideo-editingvideo-generation
/** A seconds projection of speech evidence for author inspection. Missing times stay missing. */
export type TranscriptWord = {
readonly text: string;
readonly start?: number;
readonly end?: number;
};
export type FrameWords = {
readonly active: readonly TranscriptWord[];
readonly context: readonly TranscriptWord[];
};
const SAMPLE_RATE = 16_000;
const seconds = (sample: number | undefined): number | undefined =>
sample === undefined ? undefined : Number((sample / SAMPLE_RATE).toFixed(3));
/** Project exact sample evidence into the stable human/file view consumed by author tools. */
export function transcriptDocumentFromEvidence(
evidence: {
readonly sampleFrames: number;
readonly passages: readonly {
readonly startSample?: number;
readonly endSampleExclusive?: number;
readonly words: readonly {
readonly text: string;
readonly startSample?: number;
readonly endSampleExclusive?: number;
readonly score?: number;
}[];
}[];
},
identity: { readonly source: string; readonly language: string },
) {
return {
format: "hypit.transcript@1" as const,
...identity,
audio_seconds: Number((evidence.sampleFrames / SAMPLE_RATE).toFixed(3)),
passages: evidence.passages.map((passage) => {
const words = passage.words.map((word) => ({
text: word.text,
...(seconds(word.startSample) === undefined ? {} : { start_seconds: seconds(word.startSample) }),
...(seconds(word.endSampleExclusive) === undefined ? {} : { end_seconds: seconds(word.endSampleExclusive) }),
...(word.score === undefined ? {} : { score: word.score }),
}));
return {
text: words.map((word) => word.text).join(" "),
...(seconds(passage.startSample) === undefined ? {} : { start_seconds: seconds(passage.startSample) }),
...(seconds(passage.endSampleExclusive) === undefined ? {} : { end_seconds: seconds(passage.endSampleExclusive) }),
words,
};
}),
};
}
function assert(condition: unknown, message: string): asserts condition {
if (!condition) throw new Error(message);
}
export function parseTranscriptWords(value: unknown, subject = "transcript"): readonly TranscriptWord[] {
const raw = value as { readonly format?: unknown; readonly passages?: unknown } | null;
assert(raw?.format === "hypit.transcript@1" && Array.isArray(raw.passages), `${subject}: expected hypit.transcript@1`);
return raw.passages.flatMap((passage: { words?: unknown }) => {
assert(passage !== null && Array.isArray(passage.words), `${subject}: each passage needs words`);
return passage.words.map((word: Record<string, unknown>) => {
assert(word !== null && typeof word.text === "string", `${subject}: each word needs text`);
const start = word.start_seconds;
const end = word.end_seconds;
assert(start === undefined || (typeof start === "number" && Number.isFinite(start) && start >= 0), `${subject}: invalid word start`);
assert(end === undefined || (typeof end === "number" && Number.isFinite(end) && end >= 0), `${subject}: invalid word end`);
assert(start === undefined || end === undefined || end >= start, `${subject}: word end precedes start`);
return { text: word.text, ...(start === undefined ? {} : { start }), ...(end === undefined ? {} : { end }) };
});
});
}
export function wordsAt(words: readonly TranscriptWord[], at: number): FrameWords {
const active = words.filter((word) => word.start !== undefined && word.end !== undefined && word.start <= at && at < word.end);
let anchor = active.length > 0 ? words.indexOf(active[0]!) : -1;
if (anchor < 0) {
let distance = Infinity;
for (const [index, word] of words.entries()) {
const time = word.start ?? word.end;
if (time !== undefined && Math.abs(time - at) < distance) { anchor = index; distance = Math.abs(time - at); }
}
}
const last = active.length > 0 ? words.indexOf(active.at(-1)!) : anchor;
return { active, context: anchor < 0 ? [] : words.slice(Math.max(0, anchor - 3), last + 4) };
}
function searchable(text: string): string {
return text.normalize("NFKC").toLowerCase().replace(/[\p{P}\p{Z}\s]/gu, "");
}
/** Find whole, consecutive transcript words; punctuation and case do not affect matching. */
export function phraseRanges(words: readonly TranscriptWord[], text: string): readonly { start: number; end: number }[] {
const query = searchable(text);
assert(query.length > 0, "--around needs a word or phrase");
const matches: { start: number; end: number }[] = [];
for (let first = 0; first < words.length; first += 1) {
if (searchable(words[first]!.text).length === 0) continue;
let joined = "";
for (let last = first; last < words.length && joined.length < query.length; last += 1) {
joined += searchable(words[last]!.text);
if (joined !== query) continue;
const start = words[first]!.start;
const end = words[last]!.end;
assert(start !== undefined && end !== undefined, `The phrase ${JSON.stringify(text)} has missing boundary times; inspect its passage and choose --start/--end`);
matches.push({ start, end });
}
}
return matches;
}