How I'm using NotebookLM to power up my language learning - Steve Kaufmann

Open Forum
1

NotebookLM is a game-changer for language learners. Watch my new video to learn what makes it so good:

6 Likes
3

I already knew this method and I use it often. The sad part is that it’s limited to three audios a day. I agree 100%, the audio quality is great, and the way the man and woman interact makes the dialogues feel super real. It’s like a perfect middle step to understand native content better.

This connects a lot with the idea of comprehensible input in language learning.

So I just wanted to suggest something. It would be awesome to have “THIS KIND OF FEATURE INSIDE THE LINGQ APP”. It would make the app even more unique. I don’t know if it’s possible technically or financially, but it would be a great feature.

Having this before starting a lesson helps way more than the current “simplify lesson” option. That one’s okay, but it’s not the same .

Thank you, Mr. Steve Kaufmann, for your message, and thanks to the whole LingQ team as well.

4

I found a way to avoid the 3-audio limit per day, and it’s been working great for me. Using Google Gemini, I wondered if it could imitate and produce a dialogue like the ones in the Notebook KLM app. And honestly, it did an amazing job. It sounds really similar and you can tell it understands how those dialogues are made.

With a good prompt, I managed to generate a text very close to the app’s style, then I put it through a TTS, and it worked perfectly. Of course, the final quality depends on the TTS you use.

1 Like
5

Yes: 3 a day. However, after a couple of days I have so many audios to study, that 3 a day is more than sufficient. Of course, I could start paying for NotebookLM, but I get so much practical use out of the free version it’s not worth it.

3 Likes
6

So Steve introduced me to NotebookLM. But I can’t figure out where to get the text of the conversation. Is there a text file or must I use the LingQ importer? Recently, the LingQ importer has been really inaccurate so I hope there is a text file.

2 Likes
7

I didn’t find a way within Notebook LM myself…What I did though was to download the mp3 of the audio conversation and import it into Lingq as a lesson. Lingq will do the transcription as it generates the lesson.

1 Like
8

Could you provide some examples of your queries? And how are you choosing or configuring the TTS which I assume means text-to-speech?

9

THIS IS MY PROMPT

Requirements for Podcast Content Generation

For the text I generate to perfectly fit your YouTube channel’s needs and mimic NotebookLM’s style in a discussion podcast about a video your audience has already seen, the key points are:

1. Primary Goal: Mimic NotebookLM’s Style

  • The main goal is to replicate the style of podcasts generated by NotebookLM. This means balancing clarity, analytical discussion, and conversational flow, without being a linear narration of the video’s content.

2. Format and Structure

  • Two-voice dialogue: The text should alternate between two podcast “voices.”
  • No voice markers: Do not include “Voice 1,” “Voice 2,” “A:”, “B:”, or any letters or numbers at the start of each intervention. Use only asterisks (*) or hyphens (-) to indicate a change of voice, or let the transition be fluid without explicit markers if the TTS allows.
  • Length: The text should be long enough to generate approximately 8 minutes of audio. It’s better for it to be a bit longer so you can trim it, rather than being too short.

3. Style and Tone

  • Direct and simple language: Avoid complex vocabulary, long sentences, and literary or novelistic structures. The language should be conversational, as if two people are talking informally and naturally.
  • Non-moralizing: The podcast should not judge, give life lessons, or include moralizing comments (“we all must strive,” “that’s inspiring,” etc.). The tone should be one of objective analysis and commentary on facts and implications, as NotebookLM would do.
  • Focus on discussion, not full narration: Assume the audience has already seen the video. The goal is not to retell the entire story, but to comment, analyze, and give opinions on key aspects, decisions, and situations presented in the video.

4. Interaction and Content

  • High interaction between voices: Interventions should be fluid and connected. One voice should react to what the other says, complement ideas, ask questions that the other voice answers or expands on, and actively build the discussion. This is crucial for NotebookLM’s style.
  • Opinion and analysis of facts: Presenters should give their point of view and analyze why certain things happened, what their implications are, or how situations are perceived. For example, if a character is “innocent” or “vulnerable,” the dialogue should revolve around that and how it manifests in the story.
  • Use of concepts or metaphors (if direct and clear): If concepts or metaphors (like “armor” or “innocence/vulnerability”) are used to explain a character’s behavior or transformation, they should be introduced naturally and explained simply within the conversation, in NotebookLM’s style.
  • Open-ended questions at the end: Include one or two questions at the end to invite audience reflection, closing the discussion in a way that generates interaction, very much in line with NotebookLM’s closings.

5. Errors to Avoid

  • Do not reduce length: Make sure the text is always long enough for 8 minutes of audio, regardless of style adjustments.
  • Do not use literary/novelistic language: Avoid any phrase or word that sounds complex, poetic, or deviates from a direct conversational tone.
  • Do not moralize: Avoid any comment that implies a teaching, judgment, or moral about the events or characters’ actions.
  • Do not provide a simple narration/summary: The text should not be a recount of the video’s facts, but an analysis and commentary on them.
  • Do not have poor interaction: Voices should not give isolated speeches. The conversation should flow with comments, questions, and replies between presenters.

6. Output Language

  • The podcast must always be produced in Chinese, regardless of the original video’s language.

Are you ready for me to provide a link and start producing the podcast?

2 Likes
10

I made note of all this in case I decide that NotebookLM isn’t meeting my needs. Thanks!

1 Like
11

Worth upgrading? I am currently living in the USA I am getting the first month free of cost I am given an option to cancel it before subscription starts charging.

12

Ugrading what? NotebookLM? Gemini? Lingq?

13

There is a limit to generating summary audios. Three videos a day at max. They offer me to upgrade. I am talking about NotebookLM.

1 Like
14

Hi Rafael, we are looking at how to incorporate some of thse features, as well the whole question of AI to improve LingQ. Stay tuned.

5 Likes
15

You can use a vaariety of sources, copy paste a text, a URL of a test, an audio file. You may have to use Chat GPT to create a text of interest on a subject. I sometimes just import the audio I download from one or several LingQ lessons. Note you hae to convert the WAV file of the resulting audio overview into mp3. Once you have the MP3 you can import it to create a new lesson on LingQ.

3 Likes
16

This has been a game changer for me. I turn one article into a mini course with a podcast, study guide, FAQs, and briefing note.

2 Likes
17

This note uses prompts https://forum.lingq.com/t/how-im-using-notebooklm-to-power-up-my-language-learning-steve-kaufmann/1506874/9 in Gemini to generate audio.

I prefer the free version of NotebookLM. Simple, eash, and 3 a day is more than enough for me to study. After a few days I have a lot to study.

Certainly that may not meet your needs.

1 Like
18

Just wanted to share a small but cool discovery with NotebookLM.

Previously, I was looking for a way to get around the 3-audio-per-day limit in the app. What I did was use Gemini or ChatGPT to recreate something similar to the audio that NotebookLM generates, those podcast-style dialogues based on a source. In fact, in a previous post I shared a long, detailed prompt I had put together to make sure the AI would respond in the right tone and style. It worked, but it was definitely a bit tedious.

Well, turns out there’s a MUCH EASIER WAY, and it’s built DIRECTLY INTO NOTEBOOKLM.

If you go to the Questions section and ask something like:
“Could you generate a podcast-style conversation between two hosts discussing this source in depth, using the same tone and format as your own audio generation style?”
—it works perfectly.

You get a script that feels just like the native audios from the app: natural, engaging, around 8 minutes long, and closely tied to the original source. The quality is basically the same as the official audio content, only now you can generate as many as you want, without the 3-per-day limit.

Now of course, NotebookLM only gives you the text, not the audio. So to actually turn it into something you can listen to, you’ll need a good TTS (text-to-speech) engine.

And here’s the catch: for some languages, like Chinese, the default TTS in platforms like LingQ is still pretty basic. Hopefully future updates will improve this. Otherwise, we’ll still need to use external paid TTS services to get decent quality. That’s what I’m doing now. It works well, but it does take more effort, matching the text from NotebookLM with the audio from the TTS, and uploading both into LingQ as a custom lesson.

Still, it’s a solid workaround, and honestly, the quality you get from this setup is impressive.

2 Likes
19

Which paid audio TTS service do you recommend?

20

there are lots of them out there. The good ones are expensive