YouTube lecture to notes
Paste a YouTube lecture link. Capture slide bullets, formulas, diagram labels and the spoken lecture with timestamps.
Listens and watches. Two minutes per minute of video.
Free to try, no sign-up. Hour-long lectures are read in 15-minute visual windows.
Saved lecture MP4 on your machine? Reading slides from a local file runs in the command-line tool or the Mac app.
How it works
- 1
Paste the lecture URL
Paste a public YouTube lecture, seminar or course recording link.
- 2
Read slides and speech
Both mode transcribes the lecturer and reads slides in 15-minute windows.
- 3
Export study notes
Export chapters, key points, spoken lines and slide text to Markdown or JSON.
Why audio captions miss half the lecture
Lecturers point at equations, whiteboard steps and boundary terms without reading every symbol out loud. Scribiz pairs spoken explanations with slide text across 15-minute visual windows.
- 14:20Slide 09MECH 302 · HEAT TRANSFER
1D Transient Conduction
∂T/∂t = α · ∂²T/∂x²
α = k / (ρ · cp) = 1.4e-5 m²/s
Projected slide with governing PDE and constant
- 31:45WhiteboardEXPLICIT GRID · STEP 2
Forward Euler Stability
Fo = α · Δt / Δx² ≤ 0.5
Δx = 2 mm -> Δt_max = 0.14 s
Worked discretization step on classroom board
- 52:10Slide 24BOUNDARY AT X = L
Convective Wall Balance
-k (∂T/∂x)|L = h (T_s - T_∞)
Bi = hL / k = 0.08 (< 0.1)
Boundary condition slide past the 50-minute mark
- 14:20
"Plug that diffusivity constant into the governing equation up here before integrating."
- 31:45
"If your time step crosses this bound on the board, the numerical solution oscillates."
- 52:10
"Check this Biot number at the right boundary first to verify lumped capacitance."
Missing from audio: the PDE ∂T/∂t = α · ∂²T/∂x², diffusivity 1.4e-5 m²/s, the Fo ≤ 0.5 step bound and the Biot wall equation at 52:10.
- 14:20
Said"Plug that diffusivity constant into the governing equation up here before integrating."
On screen∂T/∂t = α · ∂²T/∂x²α = k / (ρ · cp) = 1.4e-5 m²/s - 31:45
Said"If your time step crosses this bound on the board, the numerical solution oscillates."
On screenFo = α · Δt / Δx² ≤ 0.5Δt_max = 0.14 s (Δx = 2 mm) - 52:10
Said"Check this Biot number at the right boundary first to verify lumped capacitance."
On screen-k (∂T/∂x)|L = h (T_s - T_∞)Bi = hL / k = 0.08 (< 0.1)
Every slide equation, whiteboard derivation and late-lecture boundary condition stays pinned to the sentence that explains it.
What lecture notes with slide text look like
- 3:04
Slide introducing Fourier heat conduction law and sign convention.
SaidNotice the minus sign here: heat flows down the temperature gradient, never up it.
On screenLecture 4: Steady-State Conductionq_x = -k (dT / dx)k = thermal conductivity (W/m*K)
- 12:22
Diagram of a two-layer composite wall in series.
SaidWhen two materials sit in series, sum their thermal resistances just like resistors in a circuit.
On screenR_total = L_1 / (k_1 A) + L_2 / (k_2 A)Boundary at x = 0: T(0) = T_hot
- 43:30
Worked problem slide at minute forty-three with insulated boundary.
SaidPlug those values into the right-wall condition before solving for the interface temperature.
On screenInsulated right wall: dT/dx = 0 at x = LT_interface = 348 K
Why plain YouTube transcripts fail on lectures
Plain YouTube transcripts only record what reaches the microphone. Whenever a lecturer points at a slide and says "substitute this rate into the second term here," an audio-only transcript leaves you with a half-sentence and none of the formula, chemical structure, axis label or bullet list on the board.
Running Both pairs the spoken lecture with the visible slide text on a single timeline, plus chapters and a summary you can review before an exam.
| In the lecture | Captions alone | Both (speech + slides) |
|---|---|---|
| Equations and derivations | Vague references like "this term cancels" | Visible formula lines plus the spoken step |
| Slide bullets and definitions | Missing whenever the lecturer paraphrases | Slide heading and bullets captured once per slide |
| Charts and table values | Omitted unless read aloud word for word | Axis labels, units and numbers read off the frame |
Built for 60-minute and 2-hour university lectures
University course recordings, seminar talks and board-review videos routinely run 50 to 90 minutes. Scribiz reads videos longer than 20 minutes in 15-minute visual windows (up to 2 hours without an account or on a free account), so derivations and summary slides near the end of an hour-long class are captured in full instead of cutting off early.
Static frames showing the same slide collapse into one timestamped scene. For talks without slides, the YouTube video summarizer uses fewer minutes; for programming courses in an IDE, use extract code from video.
Export to Obsidian, Notion, Anki or Markdown
Export Markdown for Obsidian or Notion with clickable timestamps, TXT for clean outlines, Context to paste slides and speech into ChatGPT or Claude in one block (can ChatGPT or Claude watch YouTube videos), or JSON (the Context object) to turn definitions into Anki cards.
Read our guide on taking notes from a YouTube video for workflow tips.
Limits and allowances
- Public YouTube videos only.
- Handwriting and tiny subscripts can be misread by the model. Check formulas against the timestamp.
- Without an account: 1 free video, up to 2 hours.
- A free account takes videos up to 2 hours and 30 minutes a month. Pro takes 6 hours and 600 minutes a month.
- A YouTube caption transcript uses one caption credit, or a tenth of a minute per minute of video when you have no credit left. Captions in any other job use a tenth of a minute per minute of video.
- Both mode uses two minutes per minute of video; Watch or Listen alone uses one.
Frequently asked questions
Does this capture slide text and formulas as well as speech?
Yes. In Both mode, Scribiz transcribes the lecturer and reads visible slide titles, bullet lists, formulas and diagram labels, grouping static frames so each slide appears once with its timestamp.Does it work on 60-minute or 2-hour university lectures?
Yes, up to 2 hours without an account or with a free account, 6 hours on Pro. Videos over 20 minutes are read in 15-minute visual windows so slides and derivations in the second half of a 60-minute or 90-minute lecture are not truncated.What if the lecture has no YouTube captions?
It still works. Both and Listen transcribe the audio directly and Watch reads the slides from the video frames, even when YouTube captions are disabled or inaccurate.Can I export lecture notes to Obsidian, Notion or Anki?
Yes. Every result exports Markdown, Context (formatted for pasting into ChatGPT or Claude), TXT, SRT, VTT and JSON, with chapters, key points, spoken lines and on-screen slide text.How does it handle chalkboard or tablet handwriting?
Typed slides and clear block writing read well. Small subscripts, cursive board writing and distant camera shots can be misread by the model, so click the timestamps to verify equations.
Use it from Claude, Cursor, Codex or your code
Got a video? Get the transcript.
Or


