Skip to content

Features

Transcribe. Understand. Export.

Everything Clea does between the moment you hand it a recording and the moment you have a document worth sending.

Features

Everything between a recording and a finished document

Transcription is the first step, not the product. Clea splits the conversation into labelled speaker turns, timestamps every line so clicking a sentence jumps the audio to it, then reads the whole transcript to write a summary, pull out action items and decisions, and mark chapter boundaries. All of it stays editable: rename a speaker once and the change follows through every line and every export.

AI transcription

Accurate speech-to-text across long recordings, with punctuation and formatting applied automatically.

Speaker identification

Clea detects who is speaking and splits the conversation into labelled turns you can rename.

Timestamped transcripts

Every line is anchored to the audio. Click any timestamp to jump straight to that moment.

AI summaries

Turn an hour of conversation into a summary, key points, decisions and action items.

Smart chapters

Long recordings are divided into chapters with their own titles, so you can navigate by topic.

Multi-language

Transcribe in dozens of languages with automatic detection, then translate the result.

Multiple export formats

PDF, Word, TXT, SRT, VTT, CSV, Markdown and JSON, each with its own layout options.

AI document generation

Transform a raw recording into a formatted report, meeting minutes or interview write-up.

Speaker diarization

Know who said what, automatically

Clea detects each distinct voice and splits the recording into labelled turns. Rename a speaker once and every mention updates — in the transcript, the summary, the insights and each export.

  • Automatic detection with no speaker count required
  • Rename, merge and split speakers
  • Reassign a single line when the model gets one wrong
  • Per-segment confidence so you know where to check
Sarah Whitfield00:12

…the single biggest drag on retention. Is that still where you'd put it?

Daniel Osei00:24

It is, and honestly it got sharper. We looked at the cohort data in March…

Sarah Whitfield00:41

Do you know where the wall is? Or is it different for everyone?

AI insights

The parts that matter, extracted

Beyond a summary: the decisions that were made, the tasks people took on, the questions asked and the topics covered — each one linked back to the moment it happened.

  • Summary, key points and topics
  • Action items with the owner where the transcript makes it clear
  • Decisions and open questions
  • Automatic chapters for long recordings

Action items

  • Send the internal write-up on the pricing migrationDaniel
  • Raise day-three import rate above 50%Daniel

Chapters

  • 00:12Retention and where churn happens
  • 00:48Narrowing to the data import step
  • 01:34Moving to usage-based pricing
Export

A document, not a text dump

Eight formats and seven document templates. Choose whether to include timestamps, speaker names, the summary, chapters and page numbers, then export something you can send without reformatting it first.

  • PDF and Word with professional layouts
  • SRT and VTT subtitles, timed to the source
  • CSV and JSON for analysis
  • TXT and Markdown for anywhere else

PDF

Professional layout

DOCX

Editable in Word

SRT

Subtitles, timed

CSV

Speaker + timestamp

VTT

Web captions

JSON

Raw structure

Before and after

A recording is data. A document is useful.

The same file, once Clea has read it, separated the speakers and written it up.

Raw recording
interview-final.mp33 min
AI-generated document
2 speakers340 words
Sarah Whitfield00:12

Thanks for making the time. I wanted to start with something you said last quarter — that onboarding was the single biggest drag on retention. Is that still where you'd put it?

Daniel Osei00:24

It is, and honestly it got sharper. We looked at the cohort data in March and about sixty percent of the churn happens in the first two weeks. People sign up, they hit a wall somewhere in setup, and they just quietly stop coming back.

Sarah Whitfield00:41

Do you know where the wall is? Or is it different for everyone?

Daniel Osei00:48

It's more concentrated than I expected. Two thirds of it is the data import step. If a customer gets their data in, they stay. If they don't, nothing else we do matters. So we've stopped trying to improve the whole funnel and we're rebuilding that one screen.

Summary

A product interview covering retention and pricing. Daniel identifies onboarding — specifically the data import step — as the dominant cause of churn, with roughly 60% of churn occurring in the first two weeks and two thirds of that traced to a single screen. The team has committed six weeks to rebuilding it against one metric. The conversation then turns to the January move to usage-based pricing, which raised revenue per account by ~18% but doubled support volume until spend caps and bill forecasting shipped.

SummaryKey pointsAction itemsChaptersPDF

What you can upload

Whatever you recorded, on whatever you recorded it

Clea reads the audio track, so the container barely matters: MP3, WAV, M4A, AAC, OGG and FLAC for audio, MP4, MOV, WebM and MKV for video. A voice memo from a phone, a meeting export, a field recording, a lecture capture — all of it arrives the same way, and there is nothing to convert first.

Audio files

MP3, WAV, M4A, AAC, OGG, FLAC and other common audio formats.

Video files

MP4, MOV, WebM, MKV and more — Clea reads the audio track.

Phone recordings

Voice memos from iOS and Android upload straight from your device.

Meeting exports

Recordings downloaded from Zoom, Teams, Meet and similar tools.

Clea works with files, not links. It does not download from YouTube, Instagram, TikTok, X or Facebook — doing so generally conflicts with those platforms' terms of service. If you have the rights to a video, save it to your device and upload it; the transcription is identical.

Your conversations contain valuable information.

Turn them into searchable, structured and professional documents with AI.