The XDA piece this week on NotebookLM and YouTube research landed on the same problem most researchers hit by mid-week. A dozen tabs of tutorials, three open lectures, a saved playlist of conference talks, and no way to recall which one mentioned that one specific framework. NotebookLM is one answer. It is not the only one, and the right pick depends on whether the bottleneck is summarisation, recall, or weaving the video sources into a wider personal knowledge base.
We tested 7 desktop apps that turn video research into searchable notes. They split into three camps: live summarisers that consume a YouTube URL and spit back a structured outline, second-brain apps that ingest the summary plus your annotations and let you query across them, and full open-source pipelines for people who want the audio, the transcript, and the embeddings to stay on local disks. We mixed and matched across all three.
What to look for in a video-to-notes app
- Accurate transcription. If the captions are wrong, the embeddings are wrong, and the search returns garbage.
- Source-grounded answers. The app should cite the exact timestamp it pulled a claim from, or it is a hallucination dressed up as a citation.
- Cross-video search. A single video summariser misses the point. The reason to ingest is to ask one question across a dozen lectures at once.
- A capture path that does not require pasting URLs one at a time. Browser extensions or watch-folder importers earn their place.
- Export that lives in plain Markdown or HTML, so the notes outlive whichever app you pick.
- Decent handling of multi-language sources. Most research is not in English.
- A clear story on privacy. Either the audio stays local, or the vendor publishes its retention and training policy.
Quick comparison
| App | Best for | Free plan | Starting price | Standout |
|---|---|---|---|---|
| NotebookLM | Source-grounded Q&A on a curated set | Generous free tier | Free, paid AI Pro from around 20 USD per month | Audio overviews, podcast-style summaries |
| Recall | Cross-source semantic search | Free with limits | About 7 USD per month | Knowledge-graph style linking |
| Reflect | Personal knowledge base with AI | 14-day trial | Around 10 USD per month | Daily-notes-first workflow with embeddings |
| Glasp | Highlight-based capture from YouTube | Free | One-time pro upgrade | Social highlight feed, community insights |
| Eightify | Fastest YouTube summariser | Free with a query cap | Around 5 USD per month | Inline summaries inside YouTube tabs |
| Mem | Auto-organising AI notes | Free tier | Around 10 USD per month | Auto-tagging across notes and clips |
| Obsidian + Smart Connections | Local-first, fully owned | Free | One-time community plugins | Markdown vault, on-device or BYO LLM |
1. NotebookLM, best overall
Google’s NotebookLM is the easy default for a reason. Drop in YouTube URLs, PDF reading lists, Drive folders, and the assistant builds a per-notebook RAG index that answers questions and cites the exact source. The Audio Overview feature is the new XDA workflow: pipe a handful of research videos in, ask for an explainer, and listen back on a walk.
Where it falls short: notebooks are isolated. There is no global search across them, which makes a 30-notebook research library awkward. The two-video daily limit on Cinematic Audio Overviews is still tight.
Pricing:
- Free tier with generous source limits.
- Google AI Pro adds expanded notebooks, longer videos, and faster generations from around 20 USD per month.
Platforms: Web app on Windows, macOS, Linux. iOS and Android companions.
Download: NotebookLM
Bottom line: the best starting point. Pair it with one of the local-first picks below once the research library outgrows the notebook model.
2. Recall, best cross-source semantic search
Recall indexes anything you save with one keyboard shortcut. YouTube videos get transcribed, articles get parsed, podcasts get embedded, and a knowledge graph stitches the related concepts together. Asking “what did Andrej Karpathy say about training data quality across the last five videos” gets a sourced answer with the right timestamp inside Recall’s player.
Where it falls short: the desktop app is electron-heavy and the iOS capture flow is slower than the browser extension.
Pricing:
- Free tier for a small library.
- Paid plans start around 7 USD per month for unlimited saves.
Platforms: Web app on Windows, macOS, Linux. Browser extension, plus iOS and Android.
Download: Recall
Bottom line: the strongest pick when the bottleneck is finding which video said what, six months after watching it.
3. Reflect, best knowledge base with AI
Reflect is a daily-notes-first app that grew embeddings and YouTube import as bolt-ons. Paste a video link, Reflect transcribes and ingests it, then surfaces relevant passages while you write. The mental model is closer to Roam Research with an LLM than to a video summariser, which suits long-running research projects.
Where it falls short: the free tier is a 14-day trial, then it is paid only. Multi-device sync is fast but the editor sometimes lags on a 5,000-note vault.
Pricing:
- 14-day free trial.
- Around 10 USD per month.
Platforms: Windows, macOS, web, plus iOS and Android.
Download: Reflect
Bottom line: pick this when the research is feeding a writing pipeline, not just a recall problem.
4. Glasp, best highlight-first capture
Glasp turns YouTube into the same kind of source you would highlight in a book. Mark passages of the transcript while watching, attach a note, and the highlights pile up in a personal feed. The social side, where the same passage gets highlighted by other researchers, is the surprise. The desktop browser extension is the main surface.
Where it falls short: there is no real desktop app, just the extension and web. Power-user search and tag management lag behind Recall.
Pricing:
- Free.
- Optional one-time pro upgrade for advanced AI features.
Platforms: Browser extension on Windows, macOS, Linux. Web library.
Download: Glasp
Bottom line: the best fit when the goal is curated, hand-picked passages rather than a “summarise this for me” answer.
5. Eightify, fastest YouTube summariser
Eightify is the quickest path from URL to outline. It sits as a Chrome and Edge extension, opens a panel beside the YouTube player, and emits the key timestamps with summaries in seconds. For day-to-day research clearing, it removes the temptation to watch every 40-minute lecture in full.
Where it falls short: it does not index across sources. The summary is per video, then it is up to you to file it.
Pricing:
- Free with a small daily query cap.
- Paid from around 5 USD per month.
Platforms: Browser extension on Windows, macOS, Linux.
Download: Eightify
Bottom line: the right utility to keep open in the browser while triaging a research backlog. Pair it with NotebookLM or Recall for cross-source recall.
6. Mem, best for auto-organising AI notes
Mem is what you get when an AI sits in front of a notes app. Save a video clip and Mem auto-tags it, auto-suggests related notes, and surfaces a follow-up question that is usually annoyingly relevant. The pitch is “stop foldering your notes”, and on a research workflow with hundreds of clips it lands.
Where it falls short: the YouTube import flow is a clipper, not a native ingester, so you save a transcript chunk rather than the full source. Export is improving but not yet plain-Markdown clean.
Pricing:
- Free tier.
- Mem X plans from around 10 USD per month.
Platforms: Windows, macOS, web, iOS, Android.
Download: Mem
Bottom line: pick this if you keep building elaborate folder structures and then never finding anything.
7. Obsidian + Smart Connections, best local-first
For the researcher who wants every byte on local disk, Obsidian plus the Smart Connections community plugin is the most flexible setup. Use a YouTube-to-Markdown plugin or a yt-dlp plus Whisper pipeline to land transcripts in the vault, then let Smart Connections build embeddings against a local model. The full pipeline runs offline.
Where it falls short: it is a setup project. The vendor-managed apps above are working in minutes. This one takes a weekend, but the data stays yours forever after.
Pricing:
- Obsidian is free for personal use.
- Smart Connections and most YouTube ingestion plugins are free or pay-what-you-want.
- A local LLM runs on existing hardware or via Ollama.
Platforms: Windows, macOS, Linux.
Download: Obsidian
Bottom line: the best long-term home for research notes, especially when the research will outlive any one vendor’s pricing model.
How to pick the right one
- For the fastest setup with cited answers across a single topic: NotebookLM.
- For cross-source recall after months of saving: Recall.
- For research that feeds an active writing project: Reflect.
- For curated, hand-picked highlights: Glasp.
- For triaging a backlog of long videos before they get filed: Eightify.
- For auto-organising hundreds of clips: Mem.
- For full ownership of the notes and the model: Obsidian + Smart Connections.
Most workflows end up with two layers. A summariser like Eightify or NotebookLM in the browser, plus a long-term home like Obsidian or Recall behind it.
FAQ
Is NotebookLM free for personal research? Yes, the free tier covers most personal research workflows, with daily limits on heavy features like Cinematic Audio Overviews.
Can NotebookLM ingest a whole YouTube playlist?
Not directly. Pass individual video URLs into a notebook. A workaround is to feed transcripts in bulk via a yt-dlp pipeline.
What is the best free alternative to NotebookLM for video summarisation? Glasp for highlight-based capture and Eightify for fast summaries are both usable on free tiers. Recall has a generous free plan for cross-source search.
How accurate are AI transcripts on technical lectures? Modern Whisper-class models handle most technical content well. Accuracy drops on heavy accents, code reads, and uncommon proper nouns. Always verify a claim against the timestamp before quoting.
Can I keep my research offline? Yes, with Obsidian plus local Whisper transcription and a local LLM through Ollama or LM Studio. Nothing leaves the machine.
