Features · Transcription Search

Search every word,
spoken on camera

FrameQuery runs automatic speech-to-text with speaker diarization on every video you index, creating a searchable transcript you can query instantly.

How It Works

From raw audio to searchable transcript

FrameQuery transcribes your footage during indexing - full speech-to-text with word-level timestamps and automatic speaker separation.

Automatic transcription

Full speech-to-text runs during the indexing pipeline. Word-level timestamps mean every result lands you on the exact frame.

Speaker diarization

Multiple speakers are automatically detected and labeled. Filter search results by who said something, not just what was said.

Semantic search

MiniLM word embeddings run alongside keyword indexing. Search by meaning, not just exact words - “discussing budget” finds “we need to talk about costs”.

Search Syntax

Precise control over what you find

Query reference

"quarterly goals"

Exact phrase match. Finds the precise moment someone said this.

-interview

Exclude a term. Combine with other queries to narrow results.

@Sarah

Filter by speaker. Shows only transcript segments from that person.

codec:prores res:4k

Metadata filters to narrow by format, resolution, camera, and more.

Navigate

Click any line, jump to the frame

Click-to-seek

Click any timestamp to jump directly to that moment in the video. No scrubbing, no guessing - one click puts you on the exact frame.

Auto-scroll

The transcript highlights and scrolls in sync as the video plays. You always know exactly where you are in the conversation.

Speaker labels

Each transcript segment is tagged with a speaker label so you can clearly distinguish who is talking, even in fast-moving conversations with multiple participants.

Find Similar Lines

Pick any transcript segment and find semantically similar dialogue across your entire library. Not just keyword matching - actual meaning-based similarity powered by vector embeddings.

Export

Send transcript segments to your NLE

Select transcript segments and export them as timeline-ready files.

  • FCPXML v1.11 - Final Cut Pro
  • FCP7 XML - Legacy FCP & Premiere Pro
  • CMX 3600 EDL - DaVinci Resolve & Avid
  • LosslessCut CSV - Lossless trimming, no re-encode

FAQ

Common questions

What languages are supported?

FrameQuery uses Whisper for transcription, supporting 90+ languages with automatic language detection. You don't need to specify the language in advance.

Can I search across multiple videos at once?

Yes. Your transcript search runs across every indexed video in your library. Results are ranked by relevance and show which video and timecode each match is from.

What's the difference between keyword and semantic search?

Keyword search finds exact word matches using BM25 scoring. Semantic search uses MiniLM word embeddings to match by meaning, so “discussing finances” can find “we need to talk about the budget”. Both run simultaneously.

Does transcription cost extra?

No. Transcription runs as part of the standard indexing pipeline on every plan, including Free.