Skip to main content

Content extraction and AI research toolkit

Get the gist of any video, podcast, document or web page, with every point cited

Transcripts, clean text and AI summaries whose every point links back to the exact second or page — for professionals, creators, marketers, researchers and developers. Paste a link or drop a file: free to start, files processed in memory, and a failed request is never charged.

A demonstration: a YouTube link is fetched, its transcript appears with timestamps, an AI summary splits it into chapters, and each chapter's timestamp jumps to its line in the transcript. Illustrative example. Every chapter and key point links back to the second or the page it came from.

The problem we solve

The substance you need is locked in a 45-minute video, a 120-page PDF, an hour of audio or a page that only renders with JavaScript. Getting it out by hand means one tool for captions, another for PDFs, a third for transcription, then a chat window that loses the timestamps and cannot say where an answer came from.

GistCite takes every one of those sources in, gives back text that keeps its structure, and lets you summarize, question and draft from it with citations to the second or the page — in the web app or from the API, under one account and one quota.

What you can do

Every tool opens directly; each also has a page that explains who it is for, how it works and what it costs.

Free to start

100 units a month on the free plan (a transcript is 1, downloads are free), no card required. Go Pro for 2,000 units and the AI tools, plus API keys for automation.

See pricing →

What's new

Added in September 2026. Each card says what the feature does and what it costs in units.

  • A transcript you can click

    Click any line to play that moment in the video above it; the line being spoken stays highlighted. Search inside the transcript, step through the matches, and copy a selection with a link to the second it starts.

    Cost: Included with the transcript: 1 unit per fetch, nothing extra.

    Open a YouTube transcript
  • Markdown, Word and PDF downloads

    Save a transcript as Markdown with Obsidian-ready frontmatter and linked timestamps, as Word, or as a page to print to PDF, with caption fragments merged into paragraphs. Summaries download the same three ways.

    Cost: Free: downloads use no units.

    Download a transcript as Markdown or Word
  • Chapters and video details

    A video's own chapters show as chips that play in the page's player and become headings in the exports. The description and publish date come along too, and the API returns them as metadata.

    Cost: Free.

    See a video's chapters
  • Your caption languages, in order

    Set up to five preferred caption languages in Settings (for example “fr, en”). When a video has several tracks, your first match is used before the video's default.

    Cost: Free, for signed-in accounts.

    Fetch a transcript in your language

How it works

The same five steps for every source, in the web app and the API.

  1. Step 1

    Add a source

    Paste a YouTube link, a TikTok or Instagram post, a web address, or drop a PDF, Office file, recording or video. Nothing is stored except the text that comes out.

  2. Step 2

    Extract

    The transcript or text comes back with its structure — timestamps, chapters, page anchors, headings and tables — ready to copy, download as Markdown, Word or PDF, or send to your AI.

  3. Step 3

    Understand

    Ask for a summary, key points and chapters, or ask a question and get an answer with citations back to the exact page or second.

  4. Step 4

    Create

    Turn the source into a post, a brief, a script, captions, subtitles or clips, and download the file.

  5. Step 5

    Automate

    Wire the tools into one run on the Workflow canvas, with a record of every step. Everything in the web app is also one API call with the same account and quota, so a chain that works by hand runs on a schedule.

Read how it works in detail, including the engines it runs on

What you can check

We publish the facts a prospect can verify instead of claims they cannot.

Open engines, named

YouTube retrieval on youtube-transcript-api and yt-dlp, documents on Docling and pypdf, transcription on Whisper, summaries on Claude. Each is named on the page that uses it.

See every engine the toolkit runs on

Data handling you can read

Files processed in memory and never stored; self-service export and deletion; TLS, hashed credentials, no ad trackers; and a security page that says which audits have not been done.

Read the security and compliance page

Prices before you click

The cost of every action is shown before it runs and reported after it in the API's quota headers. The free plan is 100 units a month with no card; failed requests are never charged, and top-up units never expire.

Compare the Free, Pro and API plans

Worked examples

Reproducible walkthroughs on public material, with the cost in units. Not customer testimonials.

Independent podcaster

From a one-hour podcast episode to a newsletter, captions and three clips

One public episode on YouTube becomes a transcript, a chaptered summary, a newsletter draft, SRT captions and three short clips, all from one upload.

Follow the walkthrough

Independent consultant

From a 120-page annual report to a cited two-page brief

A public annual report PDF becomes page-anchored Markdown, a set of extracted tables, answers with page citations, and an executive brief a client can verify.

Follow the walkthrough

Graduate student

From a twelve-lecture playlist and two textbooks to a searchable study corpus

A public lecture playlist and two public-domain textbooks become one set of transcripts, summaries and chunked JSON, ready for revision or a personal AI index.

Follow the walkthrough

Built for

Pick the page for your kind of work.

Developers and AI builders

Every tool in the web app is one REST call with the same account and quota: transcripts, documents with page anchors, chunked JSON for vector stores, web extraction and Whisper.

GistCite for developers and AI builders

Questions before you start

What does GistCite do?
It turns videos, podcasts and recordings, documents and web pages into transcripts and clean text, then summarizes them with citations to the second or the page, answers questions about them, and drafts posts, briefs, captions and clips from them. The web app and the API share one account and one quota.
Who is it for?
Independent professionals and consultants, content creators and podcasters, marketers and social analysts, researchers and students, and developers and AI builders: anyone whose work crosses videos, documents and recordings and has to say where a point came from.
What is free and what is paid?
The free plan includes 100 units a month for YouTube transcripts (with Markdown, Word and PDF downloads), document extraction, PDF editing, the Markdown editor, text to speech and dictation. Pro ($15 a month list price, 2,000 units) adds AI summaries, Ask and Create, Media Tools, Web Extractor, Sentiment Analysis, Open Library and batches. API ($49 a month list, 10,000 units) is the same at volume. Prices are charged by Stripe and shown live on the pricing page.
How is pricing structured?
By quota units. A transcript, a document or a page is 1 unit; an AI summary is 10 at Standard or 30 at Deep; transcription is priced per started minute; OCR per page. Downloads are free. The cost of an action is shown before it runs, and every metered API response reports what it cost.
Do failed requests or unused units cost anything?
A request that fails is never charged. "Fails" has a fixed meaning: the request ended with an error instead of a result, whether the cause was ours, the source's (no captions, an unreadable file) or the model's (a refusal, a timeout). Units are taken only when a result is returned. A result you are unhappy with is a different case, covered in the next answer. A plan's monthly units reset each month; top-up packs (500 units) never expire and carry over from month to month, on any plan, Free included.
What if an AI result is wrong or unusable?
AI output is not guaranteed to be right, which is why every point links to the second or page it came from, so you can check it. If a summary, answer or draft misstates its source or cites a place that does not support it, write to support within 14 days with the source and what is wrong, and we return the units that result cost. Extraction (transcripts, document text) is only as accurate as the captions or the file: auto-generated captions and scans carry their own errors.
How long does it take to get started?
Seconds. Paste a YouTube link on the homepage and the transcript appears; no account is needed for a few transcripts. An account takes an email address; Pro is a Stripe checkout.
How is my data protected?
Uploaded files are processed in memory and never stored; only extracted text is kept, in your history, and you can export or delete everything yourself. TLS in transit, hashed passwords and API keys, no advertising trackers. We describe our security controls honestly, including what we have not been audited for.
Which integrations and formats are available?
A REST API with API keys for every tool; exports to TXT, SRT, VTT, Markdown, DOCX, PDF, XLSX, CSV, JSON and more; and the same content flows between tools, so a transcript can be summarized, questioned, turned into a draft or captioned without re-uploading. Python and JavaScript SDKs, an MCP server for AI assistants, an n8n node, a Chrome extension and a Zapier app are listed on the developers page, each marked Coming soon until it is published.
Which AI models are used?
Summaries, Ask and Create run on Claude by Anthropic; transcription runs on Whisper; document extraction on Docling and pypdf. Only extracted text is sent to a model, never the original file.

Read every question, including each tool's own

Start with a transcript

Paste a YouTube link and the transcript is yours in seconds. 6 tools are free to start; Pro adds the AI, media and research tools.