Why NotebookLM caps each audio/video source at 200MB, what it means for typical lecture and seminar recordings, and the three proven methods to upload anyway — without losing quality.
NotebookLM is Google's AI research and writing assistant. It lets you upload "sources" — documents, audio, and video — that the AI grounds its answers in. Each individual audio or video source is capped at 200MB per file.
For text-based PDFs this is rarely a problem. But for the most useful kind of source — long lectures, seminar recordings, conference talks, Zoom meetings — 200MB is far below what a single recording produces.
Here is roughly how much disk a typical recording takes, encoded with the H.264 codec most cameras and screen recorders default to:
| Recording Type | Typical File Size (1 hour) | Fits in 200MB? |
|---|---|---|
| Zoom recording (720p) | 300–600 MB | Rarely |
| Lecture / seminar (720p) | 700 MB – 1.2 GB | No |
| HD lecture (1080p) | 1.5 – 2.5 GB | No |
| Conference talk (1080p) | 2 – 4 GB | No |
In practice, anyone trying to feed an entire lecture to NotebookLM hits the limit on the first upload.
There are three realistic ways to get a multi-gigabyte recording into NotebookLM. Each has a different trade-off between quality, time, and cost.
| Method | Quality | Time | Cost |
|---|---|---|---|
| Re-encode at lower bitrate (compression) | Lossy — audio degrades, image softens | 10–60 min per hour of video | Free (HandBrake) |
| Cloud transcoding service | Varies by service | Slow (upload + process + download) | Paid, often per-minute |
| Auto-split into 200MB chunks (lossless) | Lossless — bit-exact copy | Seconds to under a minute | One-time $9.99 + free trial (Notebook MV Splitter) |
Re-encoding throws away information. The audio in a lecture is the most valuable channel for the AI to transcribe — degrading it to fit a file size limit defeats the purpose of using NotebookLM in the first place.
Splitting, in contrast, copies the original streams byte-for-byte into smaller files. The audio and video stay bit-exact. You just end up with N parts instead of one big file, and upload each part as a separate source. NotebookLM treats them as one knowledge base.
The catch with naive splitting is that cuts often land in the middle of a sentence — making transcripts harder to read across the boundary. Smart splitters look for natural break points (silence + scene change) instead.
This is the workflow with Notebook MV Splitter, the desktop tool we built for exactly this use case:
Does splitting reduce my video quality?
No. Notebook MV Splitter uses lossless codec copy — every byte of audio and video is preserved exactly. The parts are bit-identical to the original.
Does NotebookLM understand split files as one source?
NotebookLM treats each upload as a separate source, but the AI grounds answers across all of them within a single notebook. So one lecture in 10 parts behaves like one knowledge base.
Does Notebook MV Splitter work on Mac?
Yes — the underlying engine runs on both Windows and macOS. The Microsoft Store build is Windows-only; macOS users can run the app directly.
Can I split many recordings at once?
Yes. Batch mode scans a whole folder, queues every video, and processes them sequentially.
What video formats are supported?
MP4, MKV, MOV, AVI, WebM, FLV, WMV, and other common formats handled by FFmpeg.
Is there a free trial?
Yes. Install from the Microsoft Store and try every feature free during the trial period. After that it is a single one-time purchase of $9.99 — no subscription, no recurring fees.
Will my video uploads stay private?
Yes. Notebook MV Splitter runs entirely offline — your files never leave your computer. The only network call is the Microsoft Store license check.
Notebook MV Splitter handles a 2GB recording in under a minute. Lossless, offline, free trial then one-time $9.99.
download Get on Microsoft Store