Google NotebookLM can extract a transcript from a Youtube video

i converted all the 4150 html files into a pdf files
which took a while

then i combined them into one pdf

Code:
qpdf --empty --pages pdf_manpages/*.pdf -- freebsd_manpages.pdf

that created a 577mb pdf
which is too large to upload to google notebooklm which has an upload filesize limit of 200mb

so i split the pdf into 4 pdfs using the following command

Code:
qpdf freebsd_manpages.pdf --split-pages=4000 freebsd_manpages_part_%d.pdf

that created the following files

Code:
  -rw-r--r-- 1 djwilcox djwilcox 130M Mar 26 22:31 freebsd_manpages_part_00001-04000.pdf
  -rw-r--r-- 1 djwilcox djwilcox 161M Mar 26 22:31 freebsd_manpages_part_04001-08000.pdf
  -rw-r--r-- 1 djwilcox djwilcox 187M Mar 26 22:31 freebsd_manpages_part_08001-12000.pdf
  -rw-r--r-- 1 djwilcox djwilcox 101M Mar 26 22:31 freebsd_manpages_part_12001-14361.pdf

which are less than the 200mb upload limit

just uploading the files to google drive
then ill create a public link for the files

and after that upload them to notebook and try querying them
 
are you referring to the manpages
and if its possible to convert them to another format instead of pdfs

i converted the man pages to html
so you could convert the html to epub files instead using pandoc

if you wanted to read them on a phone

using pdfs with google notebooklm is the easiest option

you can give notebook a url but it only crawls that page and doesnt follow any links
so you can give a url and get it to crawl a website for example

so its actually easier to save webpages as pdf and upload them as a source
however the pdfs are actually rendered as plain text

you can upload plain text or markdown file

ill post the links in a minute

just watching the "The Amazing Spider Man"

most people dont know the modern films are based on a real guy from the 70s
and he could really climb walls, fire webs and even do kung fu


20250326_23h49m38s_grim.jpg



20250326_23h51m33s_grim.jpg


20250326_23h49m05s_grim.jpg
 
Eureka

I uploaded the pdfs of the manpages
and the Freebsd Handbook pdf to Google NotebookLM

I then gave it a prompt asking "tell me about the fortune command"
and it worked it queried the manpage and gave me a response with links to the source

Converting man pages to pdfs for Google NotebookLM


manpages converted to 4 pdf files
each under 200mb for Google NotebookLM

Google drive folder public download link
Note you need to be signed into your Google account to download the folder

let me know if there are any issues downloading the files


Freebsd handbook - download the pdf to use as a source


login to notebooklm


and create a new notebook called Freebsd

then upload the manpages pdfs and the Freebsd handbook pdf as sources

then you can query the handbook and manpages with one prompt, very cool

20250327_00h12m14s_grim.png


20250327_00h12m26s_grim.png
 
Really nice. I hope that NotebookLM will help me to focus my attention only on the parts that I really need to accomplish a project. Furthermore usually I understand the concepts by reading a lot of pratical examples. For me to study only the theory is not enough. So,if I ask to NotebookLM to offer to me some example of structured command,would it do it ?
 
the advantage is that NotebookLM only queries the sources ( pdf files ) you give it
like the manpages and handbook

rather than trying to search the web and coming up with nonsense

you can save webpages as pdf and upload those as well to use as sources

yes it should be able to offer examples based on the sources
 
im just looking at if i can make my Notebook Public or share it

what would be cool is to have community Freebsd Notebook people could query
could be useful to new users and save the mods on the forum dealing with noob queries
 
doesnt look like thats possible with a standard google account

you only have options to share the Notebook with people in your contacts
and not create a public Notebook
 
tell me how i can use the du command to show the filesize of a folder

Markdown (GitHub flavored):
You can use the `du` command along with specific options to display the filesize of a folder.

To show the total size of a folder in a human-readable format, you can use the following command:

```sh
du -sh <folder_path>
```

Where `<folder_path>` is the path to the directory you want to check.

Here's a breakdown of the options used:

*   `-s`: This option will **display only a grand total** for the specified directory. Without this, `du` would show the usage for each file and subdirectory within the given folder.
*   `-h`: This option will display the sizes in a **human-readable format**, using unit suffixes such as Byte, Kilobyte, Megabyte, Gigabyte, Terabyte, and Petabyte. This makes the output easier to understand.

For example, to check the size of a folder named "mydirectory" in your current location, you would run:

```sh
du -sh mydirectory
```

The output would be a single line indicating the total size of the "mydirectory" folder. For instance:

```
4.0K    mydirectory
```

This output shows that the total size of "mydirectory" is 4.0 Kilobytes.

When you copy the response from the prompt it copies the text as markdown
as shown above

below is the response in NotebookLM


20250327_00h57m03s_grim.png
 
Freebsd Mind map

after you have uploaded the sources
you will see a button above the prompt called Mind map

click the Mind map button and it will create a new button in the right column

click the button saying "Freebsd Handbook click to open the Mind map"

and this is what you will see

20250327_01h04m20s_grim.png


20250327_01h04m34s_grim.png


20250327_01h05m05s_grim.png
 
You can click items in the Mind map
and it will then open the section in the sources in the middle pain

for example if i click "User and Basic Account Management"
it then opens that section in the handbook in the middle pane

20250327_01h16m47s_grim.png


20250327_01h17m01s_grim.png
 
This actually looks pretty nice for documentation... I'm still plodding along with Plasma Wayland, compiling, recompiling, trying to figure out what's missing...

Makes me wonder what will happen if the Foundation catches wind of this... 🤤 It does take brains to put something like this together.
 
imagine if you could take the database from this forum and use it as a source in NotebookLM
then you could query the forum in addition to the handbook and man pages with a prompt

that would be really useful
 
it would be like a giant Freebsd suppository
i mean repository

sorry i cant resist any opportunity for toilet humour
being British its an integral part of our culture
 
it would be like a giant Freebsd suppository
i mean repository

sorry i cant resist any opportunity for toilet humour
being British its an integral part of our culture
You Brits have absolutely nothing on the Russian and Ukrainian sense of toilet humor, though.

Sometimes you just have to know the language, because when translated, the joke loses 90% of its punch, and it takes a lengthy explanation before the joke even makes sense.
 
You can use a url for a webpage as a source
but if the webpage is updated the source isnt updated in notebooklm

When you import a web URL (or other source), NotebookLM creates a static copy of the content at that moment.

This means it takes a snapshot of the webpage's text.

Therefore, if the original website is updated, NotebookLM's copy does not automatically reflect those changes.
 
You can use a url for a webpage as a source
but if the webpage is updated the source isnt updated in notebooklm

When you import a web URL (or other source), NotebookLM creates a static copy of the content at that moment.

This means it takes a snapshot of the webpage's text.

Therefore, if the original website is updated, NotebookLM's copy does not automatically reflect those changes.
I guess this explains the spike in connections (and subsequent outages/delays) that CloudFlare is seeing these days. It sometimes reports that a site is offline, when in fact it's not.

I think that it's basically NotebookLM's fault that I'm seeing a spike in complaints from specifically CloudFlare about being unable to make the basic HTTP connection. Did CloudFlare sign a contract with Google about prioritizing NotebookLM's connections (at expense of others, even if it is for the purpose of ML mechanics/logistics), or is it just an unfortunate fallout of the system design? Probably next to impossible for rank-and-file users like us to confirm for sure.
 
i did come across this open source version




Google notebooklm uses MeloTTS for text to speech


i tried to install MeloTTS in a jail but couldnt get it to work

these are the requirements


Code:
txtsplit
torch
torchaudio
cached_path
transformers==4.27.4
num2words==0.5.12
unidic_lite==1.0.8
unidic==1.1.0
mecab-python3==1.0.9
pykakasi==2.2.1
fugashi==1.3.0
g2p_en==2.1.0
anyascii==0.3.2
jamo==0.4.1
gruut[de,es,fr]==2.2.3
g2pkk>=0.1.1
librosa==0.9.1
pydub==0.25.1
eng_to_ipa==0.0.2
inflect==7.0.0
unidecode==1.3.7
pypinyin==0.50.0
cn2an==0.5.22
jieba==0.42.1
gradio
langid==1.1.6
tqdm
tensorboard==2.16.2
loguru==0.7.2

as you can see it uses torch and torchaudio

what im wondering is its possible to install the Freebsd pkgs
for python torch and torchaudio

Code:
Yes Master ? pkg search torch
py311-facenet-pytorch-2.5.3    Pretrained PyTorch face detection and recognition models
py311-pytorch-2.5.1_1          PyTorch: Tensors and dynamic neural networks in Python
py311-pytorch-lightning-2.3.3  Lightweight PyTorch wrapper for ML researchers
py311-pytorchvideo-0.1.5_2     Video understanding deep learning library
py311-torch-geometric-2.6.1    Graph neural network library for PyTorch
py311-torchaudio-2.2.1_6       PyTorch-based audio signal processing and machine learning library
py311-torchmetrics-1.6.0       PyTorch native metrics
py311-torchvision-0.20.1       PyTorch: Datasets, transforms and models specific to computer vision
pytorch-2.4.1_2                Tensors and dynamic neural networks in Python (C++ library)

and create a python virtual env with this option

Code:
--system-site-packages

to use the Freebsd torch libraries
 
i did come across this open source version




Google notebooklm uses MeloTTS for text to speech


i tried to install MeloTTS in a jail but couldnt get it to work

these are the requirements


Code:
txtsplit
torch
torchaudio
cached_path
transformers==4.27.4
num2words==0.5.12
unidic_lite==1.0.8
unidic==1.1.0
mecab-python3==1.0.9
pykakasi==2.2.1
fugashi==1.3.0
g2p_en==2.1.0
anyascii==0.3.2
jamo==0.4.1
gruut[de,es,fr]==2.2.3
g2pkk>=0.1.1
librosa==0.9.1
pydub==0.25.1
eng_to_ipa==0.0.2
inflect==7.0.0
unidecode==1.3.7
pypinyin==0.50.0
cn2an==0.5.22
jieba==0.42.1
gradio
langid==1.1.6
tqdm
tensorboard==2.16.2
loguru==0.7.2

as you can see it uses torch and torchaudio

what im wondering is its possible to install the Freebsd pkgs
for python torch and torchaudio

Code:
Yes Master ? pkg search torch
py311-facenet-pytorch-2.5.3    Pretrained PyTorch face detection and recognition models
py311-pytorch-2.5.1_1          PyTorch: Tensors and dynamic neural networks in Python
py311-pytorch-lightning-2.3.3  Lightweight PyTorch wrapper for ML researchers
py311-pytorchvideo-0.1.5_2     Video understanding deep learning library
py311-torch-geometric-2.6.1    Graph neural network library for PyTorch
py311-torchaudio-2.2.1_6       PyTorch-based audio signal processing and machine learning library
py311-torchmetrics-1.6.0       PyTorch native metrics
py311-torchvision-0.20.1       PyTorch: Datasets, transforms and models specific to computer vision
pytorch-2.4.1_2                Tensors and dynamic neural networks in Python (C++ library)

and create a python virtual env with this option

Code:
--system-site-packages

to use the Freebsd torch libraries

try without the jail...
 
Back
Top