Transcribe.soTranscribe.so

Turn lectures, podcasts, and videos into a searchable library. In any language.

Paste a YouTube link or upload audio. Ask anything across your whole library and jump to the exact second they said it. Cited answers, chapters, and export-ready subtitles included.

A simple HTTP API lets you plug the same engine into AI agents, automations, meeting bots, and your own apps. Read the API docs → Try it on ChatGPT →

No credit card required.·Pay only for what you use.·50+ languages.

Try this real transcript
44 Harsh Truths About The Game Of Life - Naval Ravikant (4K)
Chris Williamson
Contents
8 chapters · 513 sections
1Happiness Versus Success: Philosophical Reflections on Contentment, Desire, and Motivation
2Optimizing Sleep: Smart Temperature Regulation and the Foundations of Self-Esteem
3Decisive Action and Iterative Practice: Keys to Optimal Choices and Mastery
4Wealth Management: From Materialism to Value Creation and Fair Compensation
5Evaluating LLMs: Capabilities, Limitations, and Their Role in AI's Evolving Landscape
6Pathogens, Evolution, and Knowledge: How Humans Adapt and Defend
7Agency, Power, and the Individual: From Child Development to Cultural Conflict
8Unseen Trends: Media Oversights, Medical Limitations, and the Primitive State of Modern Biology
Q&A preview
Answer
Naval explains two distinct paths to happiness using the story of Alexander and Diogenes. The first path is through success—conquering the world, satisfying material needs, and getting what you want. The second path, exemplified by Diogenes living in a barrel, is simply not wanting in the first place. As Socrates said when shown luxuries: 'How many things there are in this world that I do not want.' Naval suggests not wanting something is as good as having it—both paths lead to the same destination of contentment [00:38–01:10]. He's not sure which path is more valid, noting it depends on how you define success [01:10–01:25].

Command Palette

Search for a command to run...

Powered by

OpenAIQwenMistral

The real cost of long videos

You remember the answer is in there. You just can't find it.

Long lectures and podcasts hold the exact moment you need. The summary won't show it. The timeline won't either. So you scrub, overshoot, and start the video over.

And when the video is in a language your tool barely understands, the transcript is wrong before you even start looking.

Per long video

3-hour lecture you need to study

+ 20–40 min scrubbing to find one explanation

+ Replaying the same 90-second clip three times

+ A summary that skipped the part you wanted

= You learned less than the hours you put in

Per week

5 long videos to get through

+ 20–30 min hunting for moments in each

= 1.5–2.5 hours of study time gone every week

How it works

Three steps from a recording to a searchable library

Real screenshots from a real transcript. No mockups, no marketing fluff. This is what you get.

  1. 1

    Drop in your audio

    Paste a link or upload a file. Anything you record, download, or screen-capture works.

    Paste a YouTube URL and pick the best speech-to-text model for your language
  2. 2

    Get a structured transcript

    Every transcript comes back with chapters, topic summaries, and timestamps so you can jump straight to what matters.

    Auto-generated chapters and topic summaries with timestamps
  3. 3

    Ask anything across your library

    Ask a question and get a cited answer with a timestamp. Search across every transcript in your library. Export subtitles if you need them.

    AI Q&A answer with inline timestamps and cited topic cards

No credit card required. Pay only for what you use.

Library, not a one-off transcript

Build a knowledge library you can actually search

Most tools give you one transcript at a time. transcribe.so turns every lecture, podcast, and video into a searchable, askable library that grows with you.

One searchable library across everything

Every YouTube link, lecture, and podcast joins one library. Find a quote across hours of content in seconds.

Ask anything. Jump to the second they said it.

Ask a question and get a cited answer with a timestamp. Click the citation to jump straight to the moment in playback.

Works in any language, automatically

67 languages with measured accuracy per language. We pick the right speech-to-text engine for you, so you focus on studying.

That's roughly 1.5–2.5 hours of study time back every week.

No credit card required. Pay only for what you use.

Independent benchmark

#4 of 80+ models on the public ASR leaderboard

We pick the most accurate transcription models on the public benchmark, not the ones with the biggest marketing budget. Lower bar means fewer errors per 100 words.

  1. Qwen3-ASRtranscribe.so
    Rank #4
    6.37%
  2. ElevenLabs Scribe v2
    Rank #8
    6.64%
  3. AssemblyAI U-3 Pro
    Rank #9
    6.80%
  4. Parakeet TDT v3
    Rank #15
    7.11%
  5. Google Chirp 3
    Rank #19
    7.37%
  6. Whisper large-v3
    Rank #34
    8.13%

Word error rate. Average across 8 datasets (AMI, Earnings22, GigaSpeech, LibriSpeech, SPGISpeech, Tedlium, VoxPopuli, Common Voice). Source: HuggingFace Open ASR Leaderboard, snapshot May 5, 2026.

21.6%
fewer errors

than OpenAI Whisper-large-v3 at the same throughput.

10.4%
fewer errors

than NVIDIA Parakeet, with Asian-language support Parakeet does not offer.

30 + 22
languages and dialects

Including Cantonese, Mandarin, Japanese, Korean, Arabic, Hindi.

A note from the maker

Hey, I'm Seunghun 👋

In 2023 I left Spotify to work on the problem of finding the useful 90 seconds inside a three-hour podcast. We built goodlisten.co, ran out of runway, and I went back to a desk job.

But I kept needing it myself. English was the easy part. The audio I actually cared about was harder: Korean podcasts where the host slips into English, Japanese conversations with three speakers, Spanish lectures recorded in noisy rooms. I was tired of spending two hours just to find the two minutes that mattered.

So in 2025 I stopped trying to build for “the market” and built the tool I wished existed, for one very specific user: me. If it saves you time, tell me. If it doesn't, tell me directly. That's how it gets better.

Seunghun

Who it's for

Built for learners. With an API for builders.

Whether you're studying from a 3-hour lecture, a foreign-language podcast, or building an AI agent that needs to listen, the same engine handles it.

Students and lifelong learners

Turn every lecture, podcast, and video you study into a searchable library. Get cited answers tied to the exact second they were said.

Learners studying in any language

Korean MOOCs, Japanese podcasts, Spanish talks, or English lectures. We pick the right speech-to-text engine for each of 67 languages, with measured accuracy.

Developers building AI apps

One HTTP API plus a Claude and ChatGPT MCP surface. Plug the same engine into agents, automations, voice memo apps, and meeting bots.

No credit card required. Pay only for what you use.

Pricing

Start free. Pay only for what you use.

No subscription required. Subscribe for a lower per-hour rate when you're ready.

Free

No commitment

$0
  • 3 Q&A questions per transcript (10/day)
  • All AI features — transcription, summaries, chapters, search
  • Pay-as-you-go after included quota
  • No subscription, no card required
Popular

Plus

Most popular

$12/mo
  • ~12 hours/month included
  • Unlimited Q&A per transcript (200/day)
  • Priority processing
  • AI summaries, chapters, topics, semantic search

Or pay-as-you-go from $1.05/hr (varies by model)

See full pricing comparison →

FAQ

Before you try transcribe.so

Start your library with one link.

Paste any lecture, podcast, or video. Free credits to start. See the exact cost before you confirm.

No credit card required. Pay only for what you use.