Transcript summarizer

Turn long transcripts into clear summaries.

By ReduzReduzUpdated May 11, 2026 Local-first AI summarizer

Spoken-word transcripts are dense, repetitive, and full of digressions. Reduz reads the YouTube transcript from the active video, the selected transcript text from a meeting recording paste, or any transcript-shaped content you have open in Chrome — and turns 90 minutes of dialogue into bullets, an executive recap, study notes, or an action-items list. Cleo uses it on recorded lectures. Anya uses it on industry-event transcripts. Theo uses it on conference-talk transcripts cited by papers.

Illustration of a person at a desk shaping content into structured output
Photo by Storyset on Pixabay

What you get

  • YouTube transcripts: read directly from the active YouTube watch page with timestamps preserved.
  • Pasted or selected transcript text: meeting recordings, webinars, podcasts, lectures — paste into the side panel or use selected-text mode.
  • Output styles: executive recap, action items, study notes, key quotes, full structured summary.
  • Long-transcript support: handles 60-90 minute transcripts on most providers; longer with Claude Sonnet 4.6 or GPT-5.5.
  • Local SQLite history with full-text search — find that webinar recap from three weeks ago by speaker name or topic.
  • Hosted Free (100 monthly credits, 1 per transcript) or Your own AI key unlimited with your own provider key.

Sample transcript summary

A 60-minute internal sales-team webinar transcript: product update, customer-objection discussion, three follow-up Q&A rounds, and a handoff plan.

  • The core discussion covers launch timing (week of 2026-06-08), three customer objections, and a handoff plan between sales and CS.
  • Three follow-ups matter: pricing evidence the deal team still needs, onboarding friction Anya flagged, and support readiness on the integration path.
  • Action items: marketing to draft the customer-objection FAQ, sales to send the pricing-evidence Loom by Friday, CS to update the onboarding checklist.
  • Top quote worth keeping (timestamp 34:12): "we underestimated how much of the friction was actually in step 3, not step 1" — useful for the launch retro doc.
  • Recommended export: turn this into a Markdown executive recap for the team channel, plus a separate action-items DOCX for the project tracker.

How it compares

Compared with raw transcript generators (Otter, Rev, Fireflies), Reduz starts from the transcript you already have and produces structured summary outputs rather than another long text artifact. Compared with YouTube-only summarizers like Eightify, Reduz also handles meeting transcripts, podcast transcripts, lecture transcripts, and pasted spoken-word text. Compared with broad AI assistants, Reduz keeps the transcript workflow focused — same extension, click-only permission, BYOK by default, local history.

YouTube transcript summaries with timestamps

For YouTube videos with auto-captions or uploaded subtitles, Reduz reads the transcript directly from the active watch page. Timestamps are preserved: clicking a timestamp in the summary seeks the YouTube player back to that moment. The same generated artifact saves to local history as a YouTube transcript summary alongside any bullet summaries from the same video.

Pasted meeting and webinar transcripts

For Zoom, Teams, or Google Meet recordings with auto-generated transcripts, paste the transcript into the Reduz side panel or use selected-text mode after pasting into a Chrome tab (a docs page, a temporary note). The full transcript becomes the source. Reduz produces executive recaps, action-items lists, key-quote extractions, or structured study notes depending on the output style you choose.

Lecture and conference-talk transcripts

Recorded lectures often come with transcripts attached. Cleo turns a 90-minute lecture transcript into a study outline with the main framework, key examples, and review prompts. Theo turns a conference-talk transcript into a structured brief that he can compare against the cited paper. Reduz separates the speaker's main argument from worked examples, follow-up questions, and digressions.

Output styles for spoken-word content

Spoken transcripts are different from articles — they have speaker turns, digressions, filler, and mid-sentence corrections. Reduz output styles handle this: executive recap strips filler and produces tight paragraphs; action items extracts decisions and assigned next-steps; key quotes preserves verbatim phrases with their context; study notes structures the content for revisiting later.

Long transcripts and what to expect

60-90 minute transcripts (10,000-15,000 words) work reliably on most providers in bring your own AI key. For multi-hour recordings or panel discussions with 5+ speakers, Claude Sonnet 4.6 and GPT-5.5 handle the the model can fit comfortably; smaller models may chunk. In Hosted Free mode, very long transcripts may prompt switching to bringing your own AI key. Transcript token counts run roughly 1.3-1.5x article counts for the same word length because spoken-word text is less compressible.

Frequently asked questions

Can Reduz summarize a YouTube transcript?

Yes. When transcript or caption text is available from the active YouTube watch page, Reduz can turn it into a summary, structured notes, key quotes, or an action-items list. Timestamps are preserved as clickable references that seek the YouTube player.

Can I summarize a pasted meeting transcript?

Yes. Paste the transcript into the Reduz side panel, or paste it into a Chrome tab and use selected-text mode. Common Zoom, Teams, and Meet transcript formats all work as long as the text is readable. Reduz handles 60-90 minute transcripts reliably; longer transcripts work best when you bring your own AI key with a larger-context model.

Does Reduz transcribe audio or video that doesn't already have a transcript?

No. Reduz works from existing transcript text (YouTube captions, Zoom auto-transcripts, Otter/Rev/Fireflies exports). If you need to transcribe raw audio first, use a transcription tool, then bring the transcript into Reduz for summarization.

How does Reduz handle multi-speaker transcripts?

Reduz preserves speaker structure when the transcript uses standard formats (e.g., "Speaker 1: ..."). Output styles like key quotes and action items reference specific speakers; the executive recap consolidates across speakers but maintains attribution for direct quotes.

Which provider works best for long transcripts?

For 60-90 minute transcripts, Claude Haiku 4.5 and Gemini Flash are cost-efficient. For multi-hour recordings or transcripts where nuance matters, Claude Sonnet 4.6 and GPT-5.5 produce noticeably better outputs. Switch providers in Reduz with one click — the same transcript can be re-summarized on a different model without re-pasting.

Is Reduz free?

Yes. Reduz includes 100 free credits a month. Using your own AI key removes the credit limit.

Do I need an account?

Not when you use your own AI key. An account is only needed for free credits, paid plans, or cloud backup.

Where is my data stored?

Summary history is stored in your browser. Cloud backup is opt-in and encrypted on your device before upload.

Which AI providers does Reduz support?

Reduz supports OpenAI, Anthropic Claude, Google Gemini, DeepSeek, and xAI Grok. You can also use free credits without setting up an AI account.