Every week I see a new AI presentation tool and everyone says this is it. This tool finally gets it and the next week the story changes again. So instead of guessing I decided to test them head-to-head.
In this video I'm putting Microsoft Copilot, the recent crowd favorite notebook LM and the deep research heavyweight skywork. ai in a straight comparison. Same brief, same presentation, no shortcuts.
We'll see which tool actually understands your content and which one just makes pretty slides. I'm Ram Gopal from presentationp process. com.
We help you present your ideas better whether you are on video or on stage. Here is the task. I want to create a medical grade visual presentation for undergraduate medical students on the topic pathophysiology of pneumonia.
I will evaluate the results based on these six parameters. The first one is depth of content. Now depth is extremely important for a professional medical presentation and the depth can actually be measured by the number of sources that are referred to for creating the presentation.
The second parameter is quality of visuals. You know the importance of visuals in a medical presentation. I want the visuals to be clear, accurate, labeled medical illustrations.
The third one is citation of sources. It is a norm to include a list of references at the end of the presentation on what are the sources that you've used to create your presentation. The next parameter is editability in PowerPoint.
Now when it comes to editability, there are two aspects. The first one is how editable is the output in general which is if I have a deck of say 16 slides and if I want to make changes to just two slides, I should be able to have the option to make changes only there. I don't want to regenerate the entire slide deck and that is one of the problems that I see in a lot of AI presentation tools.
The second aspect is I want the output to be available in PPTX format as well because when you make a complex presentation it is very likely that your audience will read ahead if you show the entire information all at once. So I want to be able to add some custom animation so I can reveal the information in stages. So editability in PowerPoint is very important when it comes to professional medical presentations.
The fifth one is watermark issues. This is actually a dealbreaker. I don't want my slides to carry the watermark saying that this is generated by this tool.
Now the final parameter is time taken. Here I'm not just talking about the time taken to generate the first draft by the AI tool. I'm talking about the total time taken which is the time taken to rework the slides because they are not according to your expectation.
So the total time taken is also a very important parameter for me. So we will take these six parameters and start testing the tools one by one. We'll start with Microsoft Copilot because it integrates naturally with PowerPoint and also with Office Suite.
So it is an obvious choice for most presenters when it comes to slide creation. Now before we start creating our prompt, there is a key thing that I want you to note. Whenever you write a prompt for a professional medical presentation, you need to be extremely clear and elaborate in your instructions.
You can't leave anything to chance. Now take a look at the kind of information that I have included in writing the prompt for this presentation. The role and objective is defined.
The role is you are an AI slide agent creating an educational medical grade visual presentation for undergraduate medical students on the topic pathophysiology of pneumonia. And then what is the objective of the presentation? It is conceptual understanding.
Then the audience and their level is clearly defined. Who is the target audience and what is their assumed knowledge and what should be avoided? In this case, it is oversimplification.
Then we talked about the content structure. Whatever is required is very clearly defined in a step-by-step flow. These are the seven things that need to be included in the presentation.
Then here are the instructions for the visual design. Every slide must include a purposeful explanatory illustration, not stock images. Use medical illustration style diagrams, not photorealistic images.
And what are the examples of acceptable visuals and what are the things to be avoided or clearly defined? And then what is my slideby-side expectation that is also very clearly written here. Each slide should have a slide title, two to four concise explanatory bullet points and a detailed description of the illustration should be shown on the slide.
What is being depicted? What is highlighted etc. The tone and language of the presentation is defined and also the instructions for the sources and citations is also mentioned here.
Only credible medical sources are to be used and inslide citation references are required wherever it is appropriate and provide a final reference slide listing all sources. And then the final output format should be like this. And I ask the AI tool to begin creating the presentation.
Now first let us copy the prompt and then go to PowerPoint. Click on the copilot icon and say create a presentation about and in the space provided I'm going to paste the elaborate prompt that I just copied. Now look at this.
It says the character limit is exceeded. This only tells me that copilot by nature is not designed to create these kind of highlevel professional presentations that require elaborate instructions on creating the presentation. So naturally we need to close this and give a much simpler instruction and see what comes out.
So let us just copy only the essentials like the role and objective, audience and level, visual design instructions and then let us go back to this and add those instructions here and let us say enter. Now let us see what kind of results we get. As you can see copilot is trying to generate an outline for the presentation.
So the outline is generated. It asks me for my preferred presentation style. I would like to have say some quick study.
And for my image preferences, I don't really have any of these as my preferred choice. So I'm going to just say suggested and say generate slides. This is quite unusual.
Copilot is taking a real long time to create its slides. It is almost 14 minutes now and still the work is not complete. Yes, I finally got my result and this is the result.
I'm going to say keep it. I've taken a bit of time to understand the content. Now let us analyze the output based on the six parameters we decided upon.
First depth of content. When I go to slides sort of view, you can see that the depth of content is pretty surface level. When I go here, you can see that the points covered are just touched upon and there is no depth to the explanation.
The reason for that is Microsoft copilot is not a deep research tool. It only accesses a few sources to get its information and it shows in the kind of results that it produces. Now the second parameter is the quality of visuals.
As you can see here, this is not the kind of visuals that you would want to use in a professional presentation. These are stylistic representations of the idea and they are not accurate medical illustrations. When you go into a symposium or a seminar and show these slides, people won't take you seriously in a professional setup.
Now the third requirement is the citation of sources. There is no mention of the sources from where the information is taken and this is unacceptable. Now the points where this presentation scores is one editability in PowerPoint naturally because this is made in PowerPoint and therefore if you want to add some animation you can do so.
Add transitions, you can change colors, you can change text everything is clearly and easily editable in PowerPoint. So it scores full marks there. There are no watermarks which is again a big plus.
And the final one is about the time taken. Now compared to a deep research kind of a presentation, the time taken by Microsoft copilot is lesser. But then the point is if the content itself is not up to the level expected, there is no point in getting the output any quicker than what it has delivered.
So overall it is clearly evident that Microsoft copilot is not the tool of choice for professional level medical presentations. So let us go to the next option. Now why notebook LM?
There are two reasons for it. The first one is it creates slide decks directly from the sources that you have uploaded which means there is no possibility of AI hallucination. The second reason which in fact is the main reason why it has become quite popular in the recent times is notebook LM uses Nano Banana Pro image generation model which is really state-of-the-art image generation model.
So the kind of output that you get is truly spectacular. So let us start using the tool and see what kind of results we get. Now the thing is there is a difference in the way you write your prompt for notebook LM compared to any other general AI tool.
General AI tools like chat GPT claude or Microsoft copilot use explicit contextbuilding prompting. So you tell the tools what to do and give all the information you need right inside the prompt. Notebook LM on the other hand uses source first implicit prompting model which means you need to upload your documents first then ask your questions and get the information.
So the first thing we need to do is to determine the sources that we need to upload to get our information for the presentation. To get the sources of information for my presentation content I usually use perplexity. ai.
In this case the prompt that I used for my medical presentation is this. You can see that I've asked Perplexity to list down just the web sources from where I can gather information for the presentation topic. Now Perplexity does its thing and you can see that it is scouring the web and trying to find the right sources that would help me to put together my content for my presentation.
Now it has done its job. Now the next step is to copy the link address from each one of these sources. Go to notebook, create new notebook and then paste those links here.
You do this one after another or if there is a way that you can actually extract all those links into one place. You can copy them in bulk and then you can paste them over here. The main important thing is every link has to start with https and that is something that I want you to remember.
And once it is done you hit enter. And you can see that all the links are now uploaded as sources. And then if you want you can always change the name of the notebook like this.
And once your notebook is named appropriately then all you need to do is to hit this option here called slide deck. Click on that and the slide deck gets generated. It takes a bit of time but then the result is quite nice.
You can see that this is the slide set that is generated by notebook LM. It looks beautiful. We will analyze the slide deck based on the six parameters.
So we are able to get a much better perspective of the output. Now let us go to the first parameter of evaluation which is depth of content. When it comes to notebook LM, we determine the number of sources.
When you use a free account, you get to upload up to 50 sources. And when you have a paid account, you can go up to 300 sources from where you can get the information for your presentation. Next, let us evaluate the quality of visuals.
As you can see here, the quality of visuals is topnotch because of the image generation tool that is used which is Nano Banana Pro. The third parameter is citation. Since this is a professional medical presentation, I want to have a list of resources included as a separate slide.
Let us see go to the last slide and see if we have such an option. I don't really see that option here. Now, this naturally takes me to the next key parameter which is editability in PowerPoint.
The reason why I stress so much on that specific parameter is for this reason. Now, take a look at this. If I want to add another slide with all the sources, I cannot really do that because the output from notebook LM is only in PDF.
That means I cannot add or delete any slide. And the other painful aspect is let us say in this case you can see that there are 15 slides and I like 13 of them and I don't like two of them. I need to generate the entire slide deck from the start.
Now I don't have any guarantee to say that only those two slides will change because the regenerated slide deck might be totally different and I might have a problem with two other slides and I need to keep doing this over and over again till I get to the point where the output is to my liking. Now that is a lot of time taken and very important thing is if I want to add custom animation to my presentation I cannot do that because when I try to export this PDF into PowerPoint it doesn't work and even if I use tools like cloudconvert. com to convert PDF to pptx the output is not really great.
Now take a look at this. This is PPT slides generated from notebookm and you can see that this output is all broken like this. And if you go to the next slide it is even worse.
And the next slide is like this. Now what this means is I need to spend a lot of time rebuilding the slide deck from ground up. I need to write all the text elements one by one manually.
Now that is a lot of work. And here comes the main part which I want you to remember. Even though this presentation looks really good.
This is not the presentation I actually want. Let me tell you why. If you remember the actual brief, the brief is I want to create a presentation on the pathophysiology of pneumonia.
Now here is the definition of pathophysiology. Pathophysiology is the study of the disordered functional changes in the body that occur due to disease or injury. Now take a look at the presentation that is generated.
I have slides like a persistent global burden with stalled mortality improvement which has nothing to do with the subject of the presentation. The next slide talks about the classification by acquisition which is not the information that I wanted. And if I go down I see some other unnecessary information like say the biomarker landscape and the risk stratification etc which are not the points that I wanted in the presentation which means even though the presentation looks good it is not the presentation that I want on the subject.
Now how did this go so wrong with notebook? The reason is the sources that were uploaded in the first place were not accurate. Now why were the sources not accurate?
Because we asked perplexity to give us the sources. Now I didn't have the time to go through each of the sources to see if each source is directly relevant to the presentation. So that means I need to spend a lot of time up front scouring the net trying to find the right kind of sources to take the information from.
So unless and otherwise I have a list of verified sources which I have seen practically manually one after another I don't really have much of a chance with getting the right kind of presentation using notebook and that is the chance I don't want to take for a medical presentation where my credibility might be at stake. So we naturally move to the third option which is skywork. First of all what is skywork.
ai? It is a suite of specialized AI agents that help people think, write, design, and build faster. So, this is the interface.
It has options for you to create documents, sheets, posters, websites, videos, and more. We'll only be interested in slides agent for this specific video. The first thing I'm going to do is to write my prompt here.
I have copied the prompt and I'm going to paste it here. Now, there are three modes in which you can generate your presentations. One is fast mode, second is deep mode and the third one is lightning mode.
This is a recent addition. I want to check how lightning mode works for a professional presentation. So let us check the lightning mode option.
Now when I choose lightning mode, I don't have the option to use nano banana pro image generation model. So let me hit enter and let skywork do its thing. Now skywork is thinking and from there it immediately went to generate.
So you can see that this is the outline for the presentation. Now it is asking me what template it needs to use. So I'm going to use one of the templates and say confirm.
Now I want you to note here the time that it has taken is around 2 minutes and that is really fast. The task is complete. Now let us take a look at the slides from the beginning.
Let us go to full screen playback. Now this is the slide deck that is generated. Let us take a look at the slides very quickly.
This is the title slide. It says alvlr lung infection mechanics. The second slide, third, fourth, fifth, sixth, 7th, 8th and the final one.
There are totally nine slides. Let's talk about the depth of content which is the first parameter. I see that the information presented is relevant to the topic of the presentation.
So that is a very good thing. When I compare it to the notebook LM which went all over the place, this is at least staying on point. What I see however is there is a lot of text written here.
I don't like my slides to have so much text. When you have a lot of text and no visuals, your audience will definitely go to sleep. The next thing that I note here is in terms of the quality of visuals, the visuals are very generic in nature.
They are not actual medical illustrations and that is something that I don't really want. And if I go to the last slide, do I have the list of references? No.
So most of the elements that I'm looking for in a professional medical presentation is not seen in this lightning fast mode. If you're going to make a school presentation or a college presentation, such a deck is okay because it covers all the points and it is covering at a very basic level. So people will be able to understand and appreciate.
But for a medical professional presentation, this is not the deck that I would like to use. So let me go back and this time I'm going to use the next mode which is the fast mode. So let us go to the slides agent.
Let us copy the same prompt here once again. Now as soon as I chose the fast mode, I now have the option of selecting Nano Banana Pro for generating my slides. Now I can hit enter.
Now this is the order in which the slide deck was created. This is my prompt. When I go down, you can see that first it has asked me for some additional information.
So how many pages that you actually want and then what is the page style you require? I chose the light theme. This gives me all the sources from where the information was taken and I can click on any of these and it directly leads me to that specific website from where the information is taken.
Now this is the PDF document from which the information was taken. You can see that it says Robinson Kotron pathological basis of disease and you can see that this is the same paper and you can see that the number of pages in this is,54. So it is a fairly elaborate piece of information.
You have a lot of information given from this particular source alone. Like this there are so many sources that were taken into account and you can see that this is the outline that is given for the slides and then the final slide deck was created. Now when I go to the slide deck, this is how the slide deck actually looks.
Let us go to the slideshow view. Full screen playback. This is the slide deck.
I want you to pay specific attention to the quality of visuals which is our second parameter. This is the second slide. You can see how elaborate, clear, and detailed the diagrams are.
You can see that every part of the diagram is clearly labeled, which makes it much easier for the audience to understand what I'm pointing to. And look at the first parameter, which is the depth of content. Every single slide is so correct, accurate, and deep.
And the reason for that is the number of sources that have been used and the quality of sources that have been used in putting together this presentation. When we go to the third parameter which is the citation, you can see the list of references here. The fourth one is in terms of the editability in PowerPoint and this is a very important parameter for me.
When I go here and click on the download option, I see that I can directly go to PPTX format and download the presentation. Now I am in the PowerPoint environment. I have downloaded the presentation as a PPTX format and you can see that the translation is perfect.
There is no distortion whatsoever be it of the text or of the diagrams and everything is beautifully converted. Now this gives me the opportunity to add custom animation. For example, let us say I want to present each of these images on a click.
I select each of them while holding the shift button. Then go to animations and say fade start on click. And then I can take this text box and this needs to wipe from left again on click.
All right. Now when I go to slideshow, this is how I start my presentation for this particular slide. Then after giving my context, I click and introduce the first one, the second one, the third one and the corresponding explanation.
So I have full control as if I have created this presentation entirely inside PowerPoint. And that's very important. And the next parameter is about the watermark.
Now since it allows me to download in PowerPoint, I have the option to remove the watermark if required. And even if I go to present in the skywork environment itself, I can go here and go to full screen playback. You can see that there is no watermark whatsoever.
Now look at this slide deck created by notebook LM in the corner. Every single slide has this notebook LM watermark and I can't do a thing about it because this is a PDF output. So it passes the fifth parameter as well.
Now the final parameter which is the time taken to create the presentation. First of all, it is fast mode and naturally as the name suggests, it gives me the result fairly fast around 18 to 20 minutes compared to a little longer time that a deep research mode takes. I don't have to do much of rework.
So the total time taken is much lesser than it takes for me to use any other method of creating the presentation. Now there is one issue that I see in this slide deck. It has covered pretty much all the information that is required.
But then there are two areas that I would want more emphasis to be given. The first one is I don't have enough information on how the immune system reacts and second I don't have a lot of information about the complications affecting multiple organ systems. So those are the two areas that I would have actually liked more information on.
So I would like to go to the deep research mode and see what kind of output that I would get from that generation. Once again I copy my prompt. Instead of fast this time I'm going to choose deep mode and nano banana pro is clicked on.
and then I need to hit enter. Now to save you some time, I'm showing you how the generation has been. Now this is the prompt.
Then I can see that it has used professional data. And this is where the difference lies between the fast mode and the deep research mode. In fact, I've created an Excel sheet comparing notebook LM and Skywork.
AI regarding the depth of content. The first point to note is in notebook LM the number of sources that you can access is limited to 300 even if you have a paid account. Whereas with Skywork.
ai 600 plus web sources are automatically referenced for every information deck. The second point is as I showed you deep research mode allows you to access information that is not normally seen when you want to do it manually. For example, you get to access academic papers, industry reports, news, statistical databases, government data, specialized databases, all these which would have actually taken you money to access.
So you get access to all these when you use the deep research mode. Now when it comes to reliability of information, you might think that notebook LM is more reliable because you define the sources from which the information is taken. However, when it comes to Skywalk.
ai's AI's deep research mode. You can see that it does cross referencing of multiple sources for verification. And that is the whole point of using deep research AI tools.
And even when you compare the different deep research AI tools, Skywork scores much better because it accesses 10 times more sources than competitors. And here is a key point that I want you to note. Skyworks deep research agent framework scored 82.
42 on the GA benchmark. outperforming competitors like open AI deep research and manis to claim the top spot. So even among the popular deep research AI tools like skywork, genspark, manis, skywork scores really high.
Then we go down and you can see that this is the card that is given here. It asks me how many pages you want to generate. I chose 16 to 20 pages and the page style which is light.
And then it has asked me this very important question which is which pathological aspects would you like to emphasize and this is the benefit of the deep research mode. Here I could see the points where I wanted to go deeper. One is immune response mechanisms and the second one is systemic complications.
I checked both of these and then I'm asked about my preferred illustration style. I chose detailed anatomical illustrations and then it asked if I had any additional requirements. I didn't have any.
So it went on to generate the to-do list. And now take a look at the number of sources it has referenced to. All these are the research papers and the website sources from where the information is taken.
You can see that it keeps running on and on. So many places from where the information is taken. Finally, it has given me the outline like this.
It gives me a quick synopsis of what is covered in each of the slides. And then finally, my presentation is created. Now let us take a look at how the deck looks when it is in the deep research mode.
Let us go to the full screen playback. This is the first slide. Now take a look at this.
There is a place for me to add the logo for my institution. Now since I can download this as a PPTX slide, I can actually add my logo here instead of what is shown here. And let us go further down.
And this is the next slide. Very clearly labeled, very accurate pieces of medical illustrations. And the quality of information is really good.
You can see how deep the information is. I actually went through this deck earlier and that is the reason why I'm able to comment on this. Every piece of information is spot-on, exactly what I wanted for my actual presentation.
Now let us see if I can add some animation and present this information one after another. So this is the deck. You can see that these are the slides and the translation is spoton.
So let us go here and let us do like the way that we did. I'm going to select this and then say wipe from left as we did earlier. I can see that these are different pieces.
I'm going to press Ctrl G to group them and then select all of these. Press Ctrl + G to group. Select this and this and then have them wipe from left on click.
Now when I go to slideshow, this is what I'll show just the title and then I talk about the information. Present the first illustration and the second illustration. So I can add animation the way I want.
Now let us go down and see if I actually have the references in the last page which actually is my third important parameter. You can see that it has got the citations and this is the summary flow diagram. If I go back to the information from the fast mode, you can see that this is the process that is shown.
Now when I go to the deep mode, you can see that the process is far more elaborate, far more points are touched, far more relevant points are touched. Let us say I don't like the content of this slide and I want skywork. ai AI to regenerate this slide.
All I need to do is to go to this option here called regenerate page 8. Click here and you can see that I have this option to regenerate this page. I can also customize the rewrite for this particular page.
Now contrast this to notebook len where you need to regenerate the entire slide deck. And as far as the time taken is concerned here it has taken a little bit longer than fast mode for sure but then I don't really have to do anything to further work on the slides. Of course, whenever I want to add some animation, I can do so.
And I showed you how quickly I can do that. So, finally, here is a decision tree that I would like to include so you can make your decision regarding the right kind of AI tool for your next presentation. The first question you ask yourself is what is the level of the presentation?
Is it a professional level or is it a beginner level? If your presentation is at the beginner level, then you can always use the pretty slides only AI tools. You know what I'm talking about like C-pilot and others.
But if your presentation is at a professional level then you have option to choose between one of the two. One is the source grounded AA tools like notebook LM or deep research AA tool like skywork. ai.
Now to decide between these two options you ask the next question which is do you have your own sources and prefer the free option. Now you pay either in terms of money or in terms of your time and with notebook LM you pay in terms of your time because if you don't have your sources you have seen what kind of results we get. Do you have the time to do the grunt work of going around the net and see and verify every single source that might be included as the source for notebook LM.
If you are able to do that work by all means go for notebook LM because it is a beautiful tool. It has got a very nice image generation model. But then many of the people who are at a professional level don't have the time to do their own grunt work and therefore I would choose the next option which is the sky work because here is an autonomous research agent that you can rely on and it does a phenomenal work not only of getting the right kind of sources but also getting the right kind of information from those sources and you'll be able to access some very specialized sources which you will not be able to access on your own because they are either paid or not known commonly.
If you want to know more about the slides agent option of skywork. ai, I've created a full review video on that. You can see the link right now on your screen.
You can click on the link and watch that video next. We thank Skywork. ai for sponsoring this video.
I'll see you inside that video.