Help & FAQ
The quick answers to the questions people ask most. Still stuck? Email info@usesourceweaver.com.
What does SourceWeaver do?
It turns media — YouTube channels and playlists, podcast feeds, and files you upload — into clean, organized transcripts packaged for your AI tools. Paste a link (or upload files), pick an output format, and SourceWeaver fetches captions or transcribes the audio, then compacts everything into a single downloadable bundle.
How do I start a job?
On the dashboard, paste one or more source URLs (one per line) or upload files from your computer, give the batch a collection name, choose an output format, and press Start Job. For a YouTube channel or playlist you'll first pick exactly which videos to include. You can watch the run's live log on the job page.
Captions vs. WhisperX — what's the difference?
When a YouTube video already has captions, SourceWeaver uses them — that's fast and costs no transcription credits. When a source has no usable captions (most podcasts, uploaded audio/video, or any video you've forced), it falls back to WhisperX, GPU speech-to-text that also labels speakers. WhisperX produces higher-fidelity transcripts but takes longer and is billed per minute of audio. Tick Force WhisperX to skip captions and transcribe everything from the audio.
What does a credit cost me?
Usage is credit-based — no monthly video quotas. The base cost is 1 credit per source (one video, one episode, or one uploaded file). If a source is transcribed with WhisperX, add 1 credit per minute of its audio. Captions-only sources incur just the 1-credit base. Your live balance and a per-job estimate are shown on the dashboard before you start, and you top up on the billing page.
Can I upload my own files?
Yes. Drag audio or video files (they'll be transcribed with WhisperX) or documents — PDF, EPUB, DOCX, RTF, TXT, MD — straight onto the dashboard's upload area. Documents are ingested as text with no transcription credit. Large uploads are spooled in chunks, so a flaky connection retries individual files rather than failing the whole batch.
How do I add more to an existing collection?
A collection is just the grouping name you give a job. Run a new job with the same collection name and its sources join that collection. To regenerate one combined output across everything in a collection — including sources from earlier jobs — use Rebuild collection on a finished job card; you'll see exactly which sources and how many credits before anything is charged.
Which output format should I pick?
- NotebookLM — combined Markdown files (≤500k words each), ideal as NotebookLM sources.
- RAG — JSONL of small, retrieval-sized chunks for your own vector store / RAG pipeline.
- Obsidian — a Markdown vault with a linked index (Map of Content).
- Logseq — a block-outline page set, zipped.
- EPUB — one e-book per collection for offline reading.
- Anki — tab-separated question/answer cards for spaced-repetition study.
You can re-run the same sources into a different format any time.
How long are my files kept?
Finished output is retained for 7 days on the free tier; Pro accounts can set a longer window (up to 90 days) in settings. Each job page shows when its files auto-delete, so download anything you want to keep before then. Retention is how storage cost is bounded — there's no per-account storage cap.
What about copyright?
You confirm you have the rights to process the sources you submit. Please respect creators' and publishers' terms. See our Terms and DMCA policy.