How do I get timestamped, citable quotes from audio for research?

A citable quote from audio needs four things: the exact words, who said them, the time in the recording, and a link a reader can follow to hear it. Audioscrape returns all four on every search result, across 18 million public podcast episodes, public government meetings, and any recordings you upload yourself. You can copy the timestamped link, create a share link to the moment, or embed a playable clip with a live transcript in an article.

What makes an audio quote verifiable

A paraphrase is not a citation. A transcript screenshot is not either, because the reader cannot check it. The standard that holds up is a pointer into the source recording at the second the words were spoken, next to the transcript of those words and the speaker’s name. That is what Audioscrape builds for every segment, whether you reach it through the website, an AI assistant, or the API.

How to do it with Audioscrape

On the website

1. Search for the wording or the idea at /search. Exact phrases go in quotes. Use speaker: "Name" if you know who said it and podcast: "Show" to narrow to a source.
2. Each result shows the snippet with matches highlighted, the speaker, the podcast and episode, the publish date, and a timestamp. Click the timestamp to open the episode page at that moment and hear it.
3. On the episode page, click any transcript segment to play from there, then click the share button next to it. You get a direct link to the moment and, on the Embed tab, an iframe snippet:

<iframe src="https://www.audioscrape.com/embed/episode/[ID]/segment/[TIME]"
        width="100%" height="400" frameborder="0" allow="autoplay"></iframe>

The embed plays the clip with a teleprompter-style transcript and links back to the full episode. ?duration=90 sets the clip length in seconds. Embeds work without a login, so readers of your article or newsletter can verify the quote in place.

4. For uploaded recordings in your workspace, excerpt share links work the same way; a private recording stays private, and you control who receives the link.

With an AI assistant

Connect Claude, ChatGPT or Cursor to https://mcp.audioscrape.com and ask for quotes with timestamps, for example “Find three moments where economists discussed the labor market this month and give me the quote, speaker, show, and timestamp for each”. The assistant calls search_audio and returns each segment’s speaker, timestamp_range and source.url. It can call create_share_link for a URL to a specific moment, and get_transcript when you want the passage in full context.

From code

POST /api/search returns, for every result, segment.text, segment.start_timestamp, segment.end_timestamp, segment.speaker, the episode and podcast metadata, and urls.segment, urls.episode and urls.audio. Store urls.segment as the citation. Reference: Search API.

What you get

• The transcript text of the segment, plus surrounding segments if you ask for context.
• Speaker name (with a slug to the person page) where attributed.
• Start and end timestamps to the second.
• A link that opens the recording at that point, a share link, and an embed.
• Episode title, show, publisher and publish date for the reference list.

Limits and honest caveats

• Transcripts are machine generated. Accuracy is high on clean studio audio and drops with accents, crosstalk and noise. Before you publish a quote, play the clip and check the words against the audio; the timestamp exists so you can.
• Speaker names in the public corpus are attributed by an evidence-gated classifier. Verify the speaker from the audio too, and report a mis-attribution from the person page if you find one.
• Search returns segments. A quote that spans a long exchange may need get_transcript or the episode page to capture in full.
• Episodes can be removed if the publisher takes them down; the citation link then breaks. For durable references, note the show, episode title and timestamp alongside the link.
• Free accounts include 100 searches per month. Browsing episode pages and embeds does not require an account.

Doing it without Audioscrape

For a single known episode, transcribe it with any tool that outputs timestamps and cite “show, episode, hh:mm:ss”. Readers still have to find the episode and scrub to the time. The searchable index and the embeddable clip are what make the citation checkable in one click.

Plans and limits: /pricing. Facts for AI agents: /docs/for-agents.