Every point links to its minute
The model cites a transcript line and the time comes from the transcript itself. Points without support in the text are dropped.
The extension is being prepared for the stores
Lecture notes where every point jumps to the moment it was said
Courses, LMSs, webinars and YouTube — even without captions. Notesavo pulls the key points from the transcript, answers questions with quotes and turns terms into Anki cards.
The sample works without installing or signing up.
- Timestamps come from the transcript, not the model
- Not just YouTube: course platforms and webinar rooms
- Your library stays in your browser
Click a timestamp and the recording jumps there
This is the real panel running on a prepared recording. The current point lights up as the video plays.
The demo recording and notes were prepared for this page.
Hello everyone. Today is lecture four of the course: graphs. We will see what they are, how to store them and how to walk them.
0:00 / 6:12
Notes · ready
Speech recognition · no subtitles on the page
In short
A lecture on graphs: vertices and edges, how to store a graph and how to traverse it. For real-world sparse graphs an adjacency list is almost always right. BFS finds the shortest path by number of edges, DFS finds cycles and components, and Dijkstra finds shortest paths in weighted graphs with non-negative weights.
Key points
- A graph is vertices and edges; edges can be directed and weighted.
- An adjacency matrix checks an edge in O(1) but takes n² memory.
- An adjacency list takes O(n + m) — the default for sparse graphs.
- BFS walks layer by layer with a queue and finds shortest paths by edge count.
- DFS goes deep with a stack or recursion: cycles, components, topological sort.
- Mark a vertex visited when you enqueue it, not when you dequeue it.
- Dijkstra takes the closest unprocessed vertex from a heap: O((n + m) log n).
- Dijkstra fails with negative weights — use Bellman–Ford.
Chapters
Terms
- Adjacency list
- Each vertex keeps a list of neighbours; O(n + m) memory.
- BFS
- Layered traversal with a queue; shortest path without weights.
- DFS
- Depth-first traversal with a stack or recursion.
- Dijkstra’s algorithm
- Single-source shortest paths with non-negative weights.
After the video you keep working material, not a retelling
Ask and verify
Answers come only from transcript fragments, and the fragments sit right next to the answer.
Review instead of forgetting
Terms become spaced-repetition cards and go to Anki in one file.
Into your system
A local library with transcript search, and one-click export.
- Anki
- Markdown
- Word
- Obsidian
- Notion
- SRT
- VTT
- Quizlet
- TXT
For videos, files and your own text
The “Notes” button appears over the player on course platforms, in webinar rooms and on YouTube. You can also drop your own file into the extension.
- Notes
Course platform
HLS stream, no subtitles
speech recognition
- AKMDlive 1:12:40Notes
Webinar
Live or recorded
subtitles or recognition
- NotesHow HTTPS works · 18:24
YouTube
Any video with subtitles
subtitles only
- lecture-04.mp4Notes
Your file
MP4, MP3, WebM, M4A, SRT
drop it into the extension
One kind of notes, everywhere
In short · Key points · Chapters · Terms · Q&A · Cards
Notesavo leaves DRM video alone and never bypasses protection — if a stream is encrypted, it says so.
No captions? We recognize the speech
For accessible videos without captions, Notesavo transcribes the audio. You see how many minutes it will use before it starts.
Audio only
The audio track is extracted in your browser; the video is never re-encoded.
In 10-minute chunks
The server measures the length itself and charges exactly that; audio is not kept.
Then like captions
The transcript is stitched without duplicates and becomes notes.
Pro includes up to 10 h of recognition a month. If the stream isn’t reachable, the extension listens to the tab during playback. On YouTube, captions only.
No subtitles
74-min video, no subtitles. Recognition will use ≈ 74 min of your limit (812 min left). Continue?
RecogniseListen to the tab
Who it’s for
The same notes, four different jobs.
- export: lecture-4.anki.tsv
Card 4 of 18
Why does BFS find shortest paths in unweighted graphs?
2:38AgainGot itA student before exams
- The problem
- 12 recorded lectures and three days until the exam.
- What Notesavo does
- Notes for every lecture, self-check questions and spaced-repetition flashcards.
- The result
- Cards go to Anki in one file, each linking to its minute in the lecture.
- АAnna09:12
Retention webinar summary:
- 0:26Measure by cohorts
- 1:18Look for a plateau
- 2:40Involuntary churn — up to ⅓
An employee after a webinar
- The problem
- A two-hour stream, and the team needs a summary by morning.
- What Notesavo does
- Key points with timestamps and takeaways — no retelling.
- The result
- Paste the summary into your team chat or Notion; links open the right moments.
“Rewatching creates a feeling of familiarity, not knowledge.”
0:42Study Hour, ep. 42Quote copied A researcher or journalist
- The problem
- You need an exact quote from an hour-long interview.
- What Notesavo does
- A searchable full transcript and verbatim quotes with timestamps.
- The result
- Quote and time land in your text — anyone can check the source.
0:00 Why graphs 0:47 Matrix or list 1:52 Breadth-first search 2:50 Depth-first search 4:01 Dijkstra’s algorithm 5:01 Summary
A video creator
- The problem
- Chapters and a description take an hour after editing.
- What Notesavo does
- Timestamped chapters, a short description and SRT/VTT subtitles from one run.
- The result
- The chapter block pastes into the description as is.
An honest comparison
If you only watch YouTube, a free summary built into your browser may be enough. Notesavo is for videos on course platforms and webinars, videos without captions, and notes that need to end up in Anki.
| Checked October 30, 2026 | Notesavo | NoteGPT | Eightify | NotebookLM, YouTube Ask |
|---|---|---|---|---|
| Where it works | Any player in the browser: courses, webinars, HLS streams; YouTube; your files | YouTube, Udemy, Coursera, Vimeo, Bilibili, files, PDF | YouTube only | YouTube, Google Drive, web pages, PDF (NotebookLM) |
| Videos without subtitles | Yes, speech recognition on any site without DRM | Yes, for files and YouTube | No | No |
| Flashcards and quizzes | Yes | Yes | No | Yes, free in NotebookLM |
| Anki export | Yes, with timestamps | Not stated | No | No |
| Your own API key | Yes, $14.99 lifetime | No | No | — |
| Free | 1 video a day up to 60 min from subtitles | 15 quotas a month across all features | 1 summary a day up to 30 min, English only | Free; YouTube Ask — in selected countries and languages, needs a Google account |
| Price | $4.99/month or $39/year | Pro ≈ $9–9.99/month | $9.99/month, or $4.99/month billed yearly | Free |
Also free: Gemini in Chrome and YouTube Ask can summarize YouTube videos. If a quick YouTube summary is all you need, they may be enough.
Data from the services’ websites and public catalogues as of October 30, 2026. Prices and limits change — if something is out of date, email hello@notesavo.com and we will fix it. Names belong to their owners.
What stays with you and what goes to a server
Stays in your browser
- Your notes and transcripts library
- Flashcards and review progress
- Settings and your own API key
Goes to the AI provider
- The transcript — to build notes and answer questions
- Audio — for recognition only, never kept
- Finished notes for videos YouTube confirms as public may be shared through a cache — without the transcript
Stored by us
- Your email and license record; the key only as a one-way hash
- Minute counters for limits — no video titles or text
- Encrypted sync data, if you turn sync on
Your own API key
Requests go from your browser straight to your provider — in the cloud or local with Ollama and LM Studio. Our server isn’t involved.
Pricing
The free version is available now. Pro and your-own-API-key come later — prices are shown for reference.
Free
$0
forever
- 1 video a day up to 60 minutes
- 5 questions per video
- Markdown and a local library
Pro
Coming later$4.99
a month or $39 a year
- 20 hours of video a month
- 10 hours of speech recognition
- All exports, flashcards and sync
Your own API key
Coming later$14.99
one-time
- Everything in Pro
- No Notesavo AI limits
- For your own provider or a local model
Questions
How accurate are the notes?
Every point is anchored to a specific place in the transcript: the model cites a line and we insert the timestamp ourselves. Points without support in the text are dropped, so you never get timestamps outside the video. Check any point with one click — the video jumps to that minute.
Which languages are supported?
English and Russian are first-class: that is where we test notes and speech recognition. Other languages work when the video has subtitles or the recognition provider supports them. You choose the notes language — e.g. English notes from a Russian lecture.
Does it work on courses with protected video?
If a platform encrypts its video (DRM), Notesavo does not bypass it and tells you so. If the video is a regular unencrypted stream or the player has subtitles, you get notes. If the stream is unavailable, “listen to the tab” mode recognises the audio while it plays.
What about YouTube?
On YouTube Notesavo only uses subtitles — manual or automatic — the same way the player loads them. We never download YouTube video or audio. If a video has no subtitles, it cannot be summarised.
What if a video has no subtitles?
On sites other than YouTube, Notesavo recognises speech. Before starting it shows the cost: “74-min video, no subtitles. Recognition will use ≈ 74 min of your limit.” Pro includes up to 10 hours of recognition out of 20 per month.
Where are my notes stored?
In your browser. Your notes library, transcripts and settings stay local. To build notes, transcript fragments go to the AI provider and are not stored by us; audio for recognition goes in 10-minute chunks straight to the provider and never stays on our server. Sync between devices is end-to-end encrypted — we only see ciphertext.
How does “your own API key” mode work?
Enter a key and the address of any OpenAI-compatible API — cloud providers or local models in Ollama or LM Studio. Requests go from your browser straight to the provider, our server is not involved, and there are no platform limits. The key is stored only in your browser profile.
Which browsers are supported?
Chrome, Edge, Yandex Browser, Opera, Brave, Vivaldi and Arc. The Firefox version works without tab audio capture, which Firefox does not support. Safari and mobile browsers are not supported yet.
Your next lecture comes with notes
Open the sample, click a timestamp and check a point against its source.
The sample works without installing or signing up.