Clips APIReal speech, indexed by word

Short videos with a CEFR level, a timed transcript and the second where each word is spoken.

Word to timestamp

Ask for a word, get back clips and the second it is spoken in each one. Seek straight to it.

Levelled, not raw

Every clip and every indexed word carries a CEFR level, so you can serve a beginner and an advanced learner from the same corpus.

Screened corpus

A curated channel list, automatic captions required, unsafe and low-speech clips dropped before anything reaches the index.

Metadata only

No video passes through us. You play each clip with the official embed, inside your own app.

Status

This site documents version 1 of the API. The endpoints are not deployed yet. The shape is published first so that integrators can review it before it is frozen. If a field is missing for your use case, this is the cheapest moment to say so.

The corpus behind it is live and growing daily. GET /corpus reports its current size and is the one endpoint that will answer without a key.

What it is

Vocaflare indexes short videos from a curated list of channels. For each clip we store the caption cues with their timings, and the catalogue words that appear in it with the second they are spoken. That index is the product.

The same index powers the Clips feed inside the Vocaflare app, so it is maintained because we use it, not because it is sold.

What it is not

Not a video host. Responses carry ids and URLs; playback is your embed, on the source platform.
Not a transcription service. Transcripts come from the platform's automatic captions.
Not a dictionary. Word meanings are returned as a convenience in seven languages, one short gloss each.