The source-of-truth layer for your AI tools
Turn the audio and video you own into AI-ready knowledge
SourceWeaver transcribes and organizes your recordings, uploads, and the content you have the rights to into clean, structured source documents, packaged to drop straight into the tools you already use.
No credit card to start · We never train AI on your data
Don't compete with your AI tools — feed them
NotebookLM, Obsidian, and your RAG stack are only as good as what you put in them. The hard part isn't the chat box — it's getting hundreds of hours of video and audio into clean, well-organized text first. SourceWeaver is that ingestion and preparation layer, built for scale: point it at sources you own or have the rights to — your own podcast, a course you've licensed, or public-domain archives like NASA's Artemis mission updates — and get back a tidy, formatted knowledge base ready to query.
One pipeline, six destinations
Export to the format your tool actually wants — not a generic transcript dump.
NotebookLM
Semantically chunked source docs, packed to the 500k-word (or custom) limit so a whole library fits in one notebook.
Obsidian
A linked vault with a Map-of-Content index and wikilinks between every episode — drop the folder in and go.
RAG / JSONL
Retrieval-tuned chunks with metadata, sized for embeddings — ready to index in your own vector store.
Logseq
Block-outline pages with properties and per-page links, zipped and ready to import into your graph.
Anki
Auto-generated question-and-answer cards for spaced repetition — turn a series into a study deck.
EPUB
A clean, readable e-book per collection — read a back-catalogue on your Kindle or tablet.
See real output — before you sign up
A public-domain NASA Houston We Have a Podcast (Artemis II) collection, processed by SourceWeaver. Pick a format to see exactly what you'd get.
Loading…
Real pipeline output, built from public-domain NASA audio (17 U.S.C. §105). Logseq ships as a multi-page vault and EPUB as an e-book file — shown here as a representative page. Try it on your own content →
How it works
Point it at your sources
Upload your own audio, video, and documents, or point it at a feed or channel you control — anything you own or have permission to process.
We transcribe & clean
Existing captions are reused; everything else is transcribed on a GPU, with optional speaker labels, then optionally cleaned of filler and errors.
Download in your format
Export per-episode or merge an entire library into one source document — in whichever of the six formats you need.
Built for whole libraries
Process an entire series or back-catalogue you have the rights to in one job — point at a whole channel or feed, then pick exactly which episodes to include.
Accurate, attributed transcripts
GPU transcription with optional speaker diarization to name who said what, and every chunk keeps its title, date, and source link.
Your data is never used to train AI
We don't sell, share, or train models on your content. The AI providers we use are contractually bound to the same. See our sub-processors and privacy policy.
Cleaning that respects the source
Optional passes fix homophones, filler, and false starts with validated, anchored edits — improving readability without rewriting meaning.
SourceWeaver is a tool you direct: you're responsible for having the rights to the sources you process. See our Acceptable Use policy.
Simple, credit-based pricing
Start free with signup credits — no card required. Go Pro for $30/mo for a monthly credit allotment, or top up any time with credit packs. You only spend credits on the work you actually run.
Ready to build your source of truth?
Bring the knowledge that's locked in hours of audio and video into the tools where you actually think.