Presentator - power point & images generator

[woeful sandal]

Yes, you read right! This discord bot allows you to create beatiful power point presentations from a given prompt, and with an NVIDIA GPU also automatically generates and adds images!
How it works
- The bot sends a request to the openai api with the given subject and indications in the marp markdown format
- We extract the images from the markdown and send them to the image generation api
- We generate the pdf and html files from the markdown
- We send the pdf and html files to the user
More informations on the github page. leave a if you like it!
https://github.com/Paillat-dev/presentator

[ornate swallow]

pos

[woeful sandal]

?

[heavy nacelle]

how do i install it

or use it

@woeful sandalhow do i use chatgpt to make power point presentations?

[woeful sandal]

it's not chatgpt, it's another model

[heavy nacelle]

ok

[vast hawk]

what is the spec requirement to run it ?

[woeful sandal]

none, if you don't generate images, if you do, you will have 2 options: 1 use dalle 2 (no requirements), 2 use stable diffusion (work better with a good gpu). In a couple of days i will add dalle support, just need to do a couple adjustements).

[jade flint]

Can i use it on mobile ?

[woeful sandal]

yes of course, you just need to run/install it onto a server/computer and run it one time

[steel vector]

Interesting. How does the image fetching work? Does it scrape a site like unsplash?

[woeful sandal]

No, it generates them , you can choose between dalle 2 or stable diffusion

[glossy bay]

Hey @woeful sandal, can you please guide me how you guys are building apps like these? I also have some cool ideas but don't know from where to start. I'm a good web developer but when it comes to AI I get confused. Here are my few questions

1. Do we need some kind of dataset to first train the model? Or is there any better approach?
2. Any blog/video from where I can get the guidance?

[woeful sandal]

No, openai already trained the models. A good point to start, is totry some things in the playground: https://beta.openai.com/playground .
Here you can type some prompts, and the AI will answer. But it's "only" a text completition AI. SO you cannot chat like chatgpt, but complete text. You need to give context, andd to "force" the AI to give you what you want.
For example, the prompt:

Here is a reciepe for a cake:

Will return you a reicepe for a cake.
But

How can i make a cake

Will sometimes something like:

 without eggs?
[...]

Because the ai completed your question.

For my powerpoint generating ai, the prompt is the following:

Here is a presentation with marp. It's not possible to make slides longer than 200 characters. to separate slides, 
"
---
"
 then go at the line. The presentatio should be for everybody, all technical words and concepts, explained. To add an image illustration , use ![bg left:50% 70%](a-description-of-the-image.png) at the beginning of the slide, just after \"---\". Use only .png. It's not possible to add technical images but only illustrations. The images are generated by an ai, the name of the file should be a detailed description of the image wanted.  For example \" ![bg left:50% 100%](a-man-wearing-a hat-ryding-a-bicicle.png)\" but don't need to show a person necessairly. The presentation is minimum 20 slides long. You can use bulletpoints. Use markdown formatting (titles, etc...). The presentation has also a conclusion. {indications} The subject of the is: {subject} The Language is: {language} <|endofprompt|>

Once you found the prompt you wanted in the playground, you can use the api, to do the exact same thing, but within your code. https://beta.openai.com/docs/api-reference/introduction . Personally i wrote my code in python, but you can also do it in plenty other languages.

@glossy bay ⬆️

[glossy bay]

Thanks @woeful sandal. I've few more questions, How can i find the right model and train it on my own data? Let's say I want AI to give result from my own data and gives result from internet if the answer isn't there in my data, how can I do this? Do I first need to create the dataset of my own and then train a pre-trained/new model to achieve this task?

[woeful sandal]

Training your own model with your data is simple. You can find how to do that here: https://beta.openai.com/docs/api-reference/fine-tunes . You will need to choose a base model to start on, and then have a database of text & expected completition. To choose a base model, you have a comparaision chart here: https://beta.openai.com/docs/models/gpt-3 . The models cannot directly search on the internet. If you want to do that, i think that the best way would be to tell the ai in the prompt to answer [GOOGLE: search simply, and then do the search by yourself, give the information back to the AI, that will process the data to make it human readable. If you haven't a big amount of data, I think you can just give the data directly in the prompt. If you have a lot of data, it can be interesting training a xustom model, and sometimes, in the "expected answers of the AI, just put [GOOGLE: the search that the AI should have done in this case. If you didn't undesrtand my explanations clearly, you can ask me

[woeful sandal]

@glossy bay ⬆️

[glossy bay]

Got it, thanks @woeful sandal

[sweet apex]

where do I type in the install commands?

[final notch]

Hello @woeful sandal, i can only install it on pc and proceed with using the app github? Cause i opened the link and read the requirements and installation, it’s pretty confusing! Should i do all this or only download and use the app?

[woeful sandal]

It's important to install marp, otherwise it won't work. If you want, you can skip the image generation steps, and use dalle-2 or no images at all.

[woeful sandal]

Wich steps are you not understanding?

[sweet apex]

it says bot ready, what do I do now?

[final notch]

Almost all these, am not a programer so i don’t know are these steps i should do or these are steps the installation automatically does? And did i get the app usage correctly? You feed in the info you need and the app will transfer them into a presentation right?

[woeful sandal]

Add it to your server and run /present subject

[woeful sandal]

So, first, download all the github

Then, open the terminal, right click and open terminal in teh folder you unzipped

Make sure you have python 3.9 https://www.python.org/downloads/ and pip installed https://pip.pypa.io/en/stable/installation/

Create a dicord bot and get his token here https://discord.com/developers/applications

Get an openai api key http://beta.openai.com/

Type the commands from the github

[final notch]

Let me try

[sweet apex]

How do I add it to my server?

[woeful sandal]

In the discord dev portal, you go to oauth, then url generato, you check the "BOT" box, then elect administrator, then you copy the link at the bottom and èpaste it in your google chrome

[tight mauve]

following

[sweet apex]

@woeful sandal , Hey the bot is working now, but despite adding the sd key to the image generation, it doesn't input the images within the presentation nor does it generate them separately, any advice on this?

[woeful sandal]

In can only work locally with sd, or remotely with dalle2

So try selecting dalle in the .env file

[sweet apex]

Sure thing thanks.

[sweet apex]

@woeful sandal I got this error when I tried to get it to give me a presentation with images "No connection could be made because the target machine actively refused it"

[woeful sandal]

could you give me the complete error?

And also a screenshot of what your .env looks like, WITHOUT the api keys

[sweet apex]

ok

I tried this step

Local image generation (optional, only if you use the local image generation option)
Open the file called Dev Console.cmd in the stable-diffusion-ui folder and run the following commands:
uvicorn main:app --reload

but it gives me this error about

File "<frozen importlib._bootstrap>", line 1014, in _gcd_import
File "<frozen importlib._bootstrap>", line 991, in _find_and_load
File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
File "<frozen importlib._bootstrap_external>", line 783, in exec_module
File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
File ".\main.py", line 1, in <module>
import sdkit
ModuleNotFoundError: No module named 'sdkit'

[woeful sandal]

You need to switch to beta

[sweet apex]

wdym?

how exactly do I do that?

[woeful sandal]

https://github.com/cmdr2/stable-diffusion-ui/wiki/The-beta-channel

[woeful sandal]

So, first you start sdui with the Start Stable Diffusion UI.cmd

Then you wait

Then you go to http://localhost

Then you follow the instructions in the link above and switch to beta, remember to save

Then you close the "stable diffusion ui" command line window that opened, and reopen Start Stable Diffusion UI.cmd and wait and reclose it

And then you can start the API in the Dev-Console

[sweet apex]

whats the command to do that?

because I got an error now after making it beta version

module 'xformers'. Proceeding without it.

I got this error when I tried to turn on developer console and ran the reload command for uvicorn

discord.errors.ApplicationCommandInvokeError: Application Command raised an exception: HTTPException: 413 Payload Too Large (error code: 40005): Request entity too large

I used dalle instead but now I get this error

[woeful sandal]

This is no an error, just an information

[woeful sandal]

Try just ignoring this error and waiting

[woeful sandal]

What is the subject that you gave it?

[umbral chasm]

I can install this program on windows 10?

[woeful sandal]

Of course

@umbral chasm

[sweet apex]

@woeful sandal so I struggled a little bit and it works for the most part, however I cant get it to upload more than 1 o 2 photos at maximum in the presentation, also the image is always something wayyyy off course to what I described, lastly if it managed to generate more than 2 images, right as it uploads the pdf on discord the terminal gives out an error 40005 about file to large or something. the project is amazing dont get me wrong, its jut if you have a solution to this, it would help soo soo much. thanks.

[woeful sandal]

That's strange, i never had that error.
To generate more images, you can add in the indications something like "Generate at least 1 image per slide" or "Generate at least 6 images". It won't work always, but most of the time. In the mean time, you can always get the presentation in the /data/YOUR_DISCORD_USER_ID/. Here, you will have different folders, each of them contains a presentation, you can sort them by last modification to see wich one is the new one. In there you should find the complete pdf.
Could you please tell me the number of pages of the pdfs it didn't managed to send?

[umbral chasm]

Hi, how to put my openai api key and discord bot token in the .env.example file and rename it to .env?

[umbral chasm]

Where this is file?

[woeful sandal]

Did you download the repositry from github?

[umbral chasm]

Yea, I downloaded

[woeful sandal]

so in the folder you should have a file called .env.example, open it in notepad

and add the values

[umbral chasm]

And it's all?

[woeful sandal]

If you want images, you can also set in that file images to dalle

And add obviously the bot to a discord server

[umbral chasm]

How to connect the discord bot to the folder?

[woeful sandal]

it's automatic

you only need to add the token to the file

and run the main.py file with python

[umbral chasm]

Ok

[umbral chasm]

Thanks

[woeful sandal]

Tell me if you need further help

[sweet apex]

@woeful sandal this does run with python 3.11 as well right?

because I want to use only python 3.11 and above from now onwards

also do I have to have the code running all the time for this to work? because currently I have to keep the terminal open for anyone on my discord server to make a presentation

[woeful sandal]

Yes

[woeful sandal]

I am not using python 11 because of compatibility problems with the pycord library

[woeful sandal]

I am planning to make a public version

[sweet apex]

Wdym public version? @woeful sandal

[woeful sandal]

A version that you'll be able to add do your server in one click

[sweet apex]

from discord.ext import commands, the discord.ext is underlined and unresolved, when I try to run main.py

@woeful sandal

any solution to that?

[woeful sandal]

pip uninstall discord

pip uninstall discord.py

pip uninstall py-cord

pip install py-cord

Run all the above commands, in the given order

[sweet apex]

ok

now when I run it, it says ready, crashes and then gives me this error @commands.cooldown(1, int(cooldown), commands.BucketType.guild)

that this line is disfunctional

can this run on python 3.10.9 ?

[woeful sandal]

@sweet apex waip

[woeful sandal]

Could you delete everything and reclone the repo? I changed something. (dont forget to put the correct values in the .env file

my bad @sweet apex

[sweet apex]

lemme try this rn

thanks

[woeful sandal]

You can ask in my server if you need more help

i see that you'realready in it

[sweet apex]

yes small problem now

its online and connected to my server, it also accepts the command /present. However when I ask it to make a presentation, it decides to crash

n error occured: Application Command raised an exception: FileNotFoundError: [Errno 2] No such file or directory: "./data/1072467202634154044/b'cHJvYmxlbXMtYWJvdXQtd3cy'2023-02-07-14-08-58/problems-about-ww2.pdf"

.

@woeful sandal

[woeful sandal]

Download marp for windows here https://github.com/marp-team/marp-cli/releases and extract the marp.exe file in the presentator folder. Do not run the file, just copy it in the resentator main folder

[sweet apex]

So its a marp problem?

[woeful sandal]

yeah, pretty sure it's one

[woeful sandal]

Oh no wait

[woeful sandal]

fixed (it was a compatibility problem with linux and windows directoryes names)

You can download the latest here https://github.com/Paillat-dev/presentator/archive/refs/heads/main.zip

Remember to put the marp.exe file

Sorry for the inconvenience

[sweet apex]

I did everything and it works perfectly, its just that the pretty often it doesnt add images, and sometimes, not all the time, but just sometimes gives me this error @woeful sandal

@woeful sandal also can I dm you regarding this bot, as I like it very much but would like some extra features added for my server, maybe we can work on this and make it something crazier than what it is 🙂

[sweet apex]

When it generates more than 2 or 3 pictures from dalle, it crashes 🥲. @woeful sandal

The pay load is too large, how do I change the setting for that?

Error 4005

[steel vector]

Image is for discord too large

Discord api have limit

Try generate only two images

[woeful sandal]

yeah, the problem is that it cann't send large files on discord

[sweet apex]

Can you create a compression code?

[woeful sandal]

i'll see if i can do in 512x512

directly

[sweet apex]

Thanks 😄

Can I discuss about a possible commision for another bot?

[woeful sandal]

DM

[woeful sandal]

ok, i'ts done, but you can directly change the code in the imagesGeneration.py file to look like this:

import requests
import os
import openai
async def generate(prompt, path, mode, openai_key):
    #r = requests.get(f"http://localhost:8000/generate_image?prompt={prompt}&path={path}")
    if mode == "sd": 
        r = requests.get(f"http://localhost:8000/generate_image?prompt={prompt}&path={path}")
        return "image generated"
    if mode == "dalle":
        openai.api_key = openai_key
        img = await openai.Image.acreate(
                prompt=prompt,
                n=1,
                size="512x512",
            )
        return img

[sweet apex]

Thanks 🙂

[earnest musk]

#openai-chatter

[woeful sandal]

?

What do you mean?

[steel vector]

thanks

[thin timber]

Iis just for IOS?

[woeful sandal]

Works on windows only

[thin timber]

TY

[versed compass]

@woeful sandal following your work and tried both Stable Diffusion and Dall-E for slides

One work example

If there a manner to change this message right in the code?

[woeful sandal]

yes

main.py line 184

[woeful sandal]

do you like it? Feel free to give any feedback!

[versed compass]

How about deciding in the config to avoid putting an image on the slides? Currently reconfiguring the bot

So there would be no errors when the user doesn't want images
when the code literally has:
if use_images != "No": imageint = "To add an image illustration , use ![bg left:50% 70%](a-long-detailed-description-of-the-image.png) at the beginning of the slide, just after \"---\". Use only .png. It's not possible to add technical images but only illustrations. The images are generated by an ai, the name of the file should be a detailed quite long description of the image wanted. For example \" ![bg left:50% 100%](a-man-wearing-a hat-ryding-a-bicicle.png)\" but don't need to show a person necessairly."

[woeful sandal]

set images to No in the .env file

[woeful sandal]

if use_images != "No": means that if it is equal to no it'll ignore that part of the code

USE_IMAGES=No

[versed compass]

reconfigured it to my instructions in my own language, this server didn't permit me to use my language here, also a clarification

"indication" should mean extra instructions or notes to the bot when creating the slides, for example, what language tone to use or which target audience the presentation is for

it should be "extra_instructions"

[versed compass]

Changing env to setting images to no returns this error when the bot is online: Application Command raised an exception: FileNotFoundError: [WinError 2] The system cannot find the file specified:

Meanwhile, the folder where the app is placed will still produce presentation files locally. But they are not uploaded onto the Discord channel, while the error is shown.

[woeful sandal]

No with capital N ?

[versed compass]

Same code/id, different times, different approaches to content.

[versed compass]

Ah, you mean this command is case sensitive.

[woeful sandal]

yes

[woeful sandal]

needs to be USE_IMAGES=No

[versed compass]

Done, created slides successfully. This result is only two slides 😦

[woeful sandal]

in the indications, say the presentation should be at least _ slides long

[woeful sandal]

new update! Added new themes and improved prompts

[visual jetty]

An error occured: Application Command raised an exception: TypeError: object OpenAIObject can't be used in 'await' expression

[woeful sandal]

do you always get that error or only sometimes? Can you try upgrading openai with pip install --upgrade openai?

[visual jetty]

i am getting the error every single time

[visual jetty]

already done that

before this error i am getting another error:
An error occured: Application Command raised an exception: AttributeError: type object 'Completion' has no attribute 'acreate'
i renamed the .acreate to .create in the main.py file

[woeful sandal]

No. It should stay acreate Can you try uninstalling the openai lib and reinstalling it? Looks like yours isn't up to date.

Do you have different python installations on your pc?

[visual jetty]

I'm using Python 3.10

[woeful sandal]

but do you have different versions installed?

[woeful sandal]

.

[visual jetty]

no only python 3.10.10

[visual jetty]

sorry i dont recall ever installing openai lib

[visual jetty]

i managed to generate the powerpoint

but so far unable to use stable diffusion yet

when i do this:
Open the file called Dev Console.cmd in the stable-diffusion-ui folder and run the following commands:
uvicorn main:app --reload
I am getting:
SyntaxError: invalid syntax
INFO: Stopping reloader process [3116]

[woeful sandal]

1. have you switched to beta?
2. have you installed uvicorn with pip installl uvicorn in the dev console AFTER switching to beta
3. have you copied the correct file in the stablediffusion folder?

[visual jetty]

apparently the !DOCTYPE html tag is give syntax error
<!DOCTYPE html>
^
SyntaxError: invalid syntax

[visual jetty]

1. Yes switched to beta
2. after entering pip install uvicorn, it says requirement already satisfied
3. i copied the main.py from presentator/img_gen_api/main.py in your github to the stablediffusion folder

[woeful sandal]

from wich file? Could you snd the whole output from the terminal?

also

did you close and reopen the webui 1 time after switching?

[visual jetty]

File ".\main.py", line 8
<!DOCTYPE html>
^
SyntaxError: invalid syntax

[woeful sandal]

what?

there is nothing html related in main.py

line 8 is:
#This is an API on the port 9009 that generates images from a prompt

[woeful sandal]

could you open the main.py file in any app (e. notepad) and check it looks like what you can find here? https://github.com/Paillat-dev/presentator/blob/main/img_gen_api/main.py

[visual jetty]

ah ok i get it now

so i entered uvicorn main:app --reload

now its at
Model loaded
INFO: Started server process [19220]
INFO: Waiting for application startup.
INFO: Application startup complete.

does that mean i can use the discord bot already and it will generate images in the Powerpoint?

my powerpoint is still plain text atm

managed to get a picture into the powerpoint but why is it always the man wearing a hat riding a bicycle?

i mean why is the picture prompt always a man wearing a hat riding a bicycle? shouldn't it be similar to the presentation prompt

[woeful sandal]

in the indication, tell him to include images related to the presentation

if you have anything you want/don't want, just put that in the indications.

[woeful sandal]

UPDATE
added new themes
improved large files handling, with zip files

[queen beacon]

Lets compare and see how my method is excellent for real time examples and explanations in open ai chat , using pictures, animations, ascii art, glistening/glow/etc to give you the actual animation as the explanation. ( good for schools , hospitals , etc) .. i think you took this a step further like i want to use an image generator program which is impressive nobody has done that so far. i was just about to. if you want to collaborate let me know.

[woeful sandal]

I am using https://marp.app for presentations pages layouts

[cloud hornet]

Interesting

[storm vortex]

how do u switch to beta?

[woeful sandal]

Just use dalle

It's simpler

[patent tangle]

imagine/Neoclassical Interior design for a large Villa living room8*12m,High3.2M,
detailed sofa Java Cream2 fabric simple detailed cushion large
glass windows night mood moon appears on the sky some night
with reflected porcelain, detailed wooden neoclassical door

[storm vortex]

dosent it cost credits?

[woeful sandal]

Yes, but same as the rest. Generating a presentation costs around 0.05 $ in your free credits

[storm vortex]

i put dalle on .env but it isnt generating photos what do i do

[woeful sandal]

dalle lowercase

?

[storm vortex]

yes

[woeful sandal]

then in the iinstructions, write "include 2 images"

in english

[storm vortex]

ok

[storm vortex]

where is instruction located?

[woeful sandal]

when running the /present command

you have the instructions option

[storm vortex]

it worked thx