Any video URL in.
Clean, timestamped text out.

Real AI speech recognition — not caption scraping — for TikTok, Douyin (抖音), Xiaohongshu (小红书), Bilibili (B站), Instagram Reels and YouTube. Built for LLM pipelines, content research, and subtitle generation.

No captions neededBest-in-class Mandarin ASRTimestamped sentences$0.025/min · no subscription8 languages, auto-detected

One API per platform — or one for all of them

Every endpoint takes public video URLs and returns the same clean JSON: full transcript, sentence-level timestamps, and the video's metadata. Mandarin-heavy platforms use a Chinese-optimized model that western transcription tools can't match.

How it works

1

Send video URLs

Paste links or call the API — share links, short links and bare IDs all resolve automatically. Mix platforms freely on the universal endpoint.

2

We fetch & transcribe

The video's audio track is extracted server-side and run through production ASR (Alibaba Paraformer-v2) — no cookies, logins, or browser automation on your side.

3

Get structured JSON

Full transcript, timestamped sentences (ready for SRT/VTT), plus author, description, duration and engagement stats — one uniform shape across all platforms.

Pricing

Pay only for minutes actually transcribed. Failed fetches and silent videos are never billed. No subscription, no minimum — billing runs on Apify's metered infrastructure with $5 free credit monthly.

$0.025per minute of video transcribed (YouTube: $0.01 flat per video)
WorkloadCost
One 30-second TikTok$0.0125
100 Douyin videos (~45s each)≈ $1.88
A 30-minute Bilibili lecture$0.75
1,000 YouTube transcripts$10.00

Why not just use Whisper?

You could — after you solve video URL resolution for five platforms, short-link expansion, CDN download auth, audio extraction, and Mandarin accuracy. These APIs package all of that behind one call:

The download is the hard part

Douyin, Xiaohongshu and TikTok actively rotate their video CDNs and link formats. We maintain the adapters so your pipeline doesn't break at 2am.

Mandarin that actually works

Whisper underperforms on Chinese short-video audio. We run Alibaba's Paraformer-v2 — the model that tops Mandarin ASR benchmarks — for 抖音 / 小红书 / B站 content.

Agent-ready structured output

Uniform JSON with timestamps and metadata drops straight into RAG, agents and analytics — no per-platform parsing, ever.