One API call turns TikTok, Reels and Shorts into clean JSON — the on-screen text, the transcription, and the entities that matter, deduplicated across thousands of clips. Stop building brittle scrapers and OCR glue.
Join the private beta — we'll email your API key. No spam.
# POST a video URL, get structured data backcurl https://api.sift.dev/v1/extract \
-H "Authorization: Bearer sk_live_..." \
-d '{"url": "https://tiktok.com/@user/video/123"}'# → clean, typed JSON
{
"transcript": "the best hidden ramen spot in Shibuya...",
"on_screen_text": ["Ichiran Shibuya", "open till 6am"],
"entities": [
{ "type":"place", "name":"Ichiran Shibuya", "confidence":0.94 }
],
"language": "en", "dedup_key": "a1f9..."
}
⚡
One endpoint
Send a URL, get JSON. We handle download, frame-dedup, OCR, transcription and extraction — you never touch the platforms.
◇
Deduped, not raw
The hard part isn't reading one clip — it's collapsing thousands into a clean, non-repetitive knowledge base. That's the core.
⬡
Any platform, any entity
TikTok, Reels, Shorts. Extract places, products, recipes, brands — define the schema, we fill it.