AI Slide Generation: Comparing 6 Tools with the Same Prompt Revealed Shocking Differences

@uchita_success
ЯПОНСКИЙ2 месяца назад · 06 июн. 2026 г.
294K
347
28
11
696

Суть

A comprehensive comparison of six AI slide generation tools—Claude, ChatGPT, Gemini, NotebookLM, Genspark, and Copilot—evaluating their design quality, Japanese text rendering, and editability for business use.

"If I'm making slides with AI, which one is actually the best?"

I was asked this while conducting AI training for a company. AI evolution is intense, so I decided to compare them once again.

I threw the "exact same prompt" at 6 tools I use and had them generate slides.

The 6 tools used:

Claude / ChatGPT / Gemini / NotebookLM / Genspark / Copilot

Despite the same instructions, the results were quite different. I'll look at each one, including appearance, text, structure, and ease of editing.

Verification Rules

To keep it fair, the conditions were as follows:

  • The topic is fixed to one.
  • The prompt is exactly the same. However, I provided the same detailed brief + same color scheme + same visual layout specifications to all tools = looking at "how well they can output common specs in high-quality business visuals + faithfulness to instructions."
  • Evaluation is on a 5-point scale across 4 axes.

The Actual Prompt Used

text
1# Role
2Please create 2 proposal slides for the following new business to be used in a management meeting. Prioritize a business-quality finish that communicates visually at a glance.
3
4# Scenes (2 slides to create)
5
6## Business Overview (Common premise for both)
7- Business Name: "Sukima Eigo" (Tentative)
8- Overview: A corporate English training subscription where you can talk to AI instructors starting from 5 minutes during commute or break times.
9- Target: Small to medium-sized enterprises that want to include English training in their employee benefits.
10
11## Slide 1: Competitive Positioning
12- Headline (Conclusion): We are in the "Sustainable x Corporate-Focused" gap.
13- Layout: 2-axis positioning map (matrix) + extraction logic for "why these 2 axes" on the right side.
14- Vertical Axis: Retention Rate (Bottom = Doesn't last / Top = Lasts)
15- Horizontal Axis: Ease of Corporate Implementation (Left = Effort / Right = Easy)
16- Plot:
17 - Station-front English schools (High price, low retention) -> Bottom Left
18 - Online English schools (Medium) -> Center
19 - Learning apps (Low price, no corporate support) -> Near Top Left
20 - Our "Sukima Eigo" -> Gap in the Top Right (Highlight with Accent Vermilion)
21
22## Slide 2: Implementation Roadmap
23- Headline (Conclusion): Steadily reaching 100 companies in 4 quarters.
24- Layout: Multi-stream schedule (Vertical = Stream / Horizontal = Q1-Q4 timeline / Dashed lines for unconfirmed starts)
25- Streams:
26 - Product: Q1 Verification version -> Q2 Improvement
27 - Sales: Q2 10-company pilot -> Q3 Official launch -> Q4 100 companies
28 - Customer Success: Q3~ (Dashed line start)
29
30# Overall Specifications
31- Aspect Ratio: 16:9 (Widescreen)
32- Background: Pure White
33- Style: Flat design (no shadows, no gradients), minimal, consulting document style
34- Font: Clean sans-serif like Noto Sans JP
35- Typography: Apple-style, elegant layout utilizing white space
36- All Japanese text must be displayed accurately (extremely important)
37
38# Color Palette (Strict Adherence)
39- Main Blue #1976D2: Titles, card heading bands, primary icons
40- Accent Vermilion #D84315: Only for 1-2 emphasized keywords
41- Light Blue #E8F4FB: Icon backgrounds, light fills
42- Body Dark Gray #222222: Body text
43- Sub Gray #888888: Auxiliary text
44- Border Gray #E0E0E0: Card borders, dashed lines
45
46# Application Rules
47- 1 Slide = 1 Message. Place the conclusion (fact) in one line at the top, and implications in a separate box.
48- Do not repeat the same words (summarize common items once in axes or headings).
49- Numerical sizes should be proportional to area, length, or thickness.
50- Prioritize "communicating visually at a glance" with graphs and diagrams.

Evaluation Axes

I generated two slides and evaluated them on the following axes:

  • Design
  • Text (Japanese) clarity
  • Ease of editing
  • Output speed (*)

These were the four axes. Note that output speed is just for reference since it runs in the background. (I personally don't mind it much.)

Putting the 6 Tools to the Test

Which one do you prefer? Since design is subjective, I had an AI evaluate them. (I used Claude, but I hid the tool names so it wouldn't be biased. lol)

うちた - inline image

**

1) ChatGPT

うちた - inline image
うちた - inline image
  • Score: Design 5, Text 5, Editing 4, Speed 4
  • My Impression: Good design. It drew the four quadrants beautifully and highlighted our company with a dashed frame. You can edit via natural language with high accuracy. Since there have been many GPT Images 2.0 generations lately, it might feel a bit familiar.

2) Gemini

うちた - inline image
うちた - inline image
  • Score: Design 3, Text 2, Editing 2, Speed 4
  • My Impression: There were typos in the text (e.g., "Positioning"). It added "10" to the official launch which wasn't instructed. It's been a while since the last major update, so this might be expected. I hope to see top-tier performance again in the next version.

3) NotebookLM

うちた - inline image
うちた - inline image
  • Score: Design 4, Text 3, Editing 2, Speed 3
  • My Impression: The design balance is good. The text is a bit crushed. Even when exported to PPT, the whole thing is an image and cannot be edited. (A PPT conversion prompt might help.) Like Gemini, I'm looking forward to updates.

4) Claude

うちた - inline image
うちた - inline image
  • Score: Design 4, Text 4, Editing 4, Speed 2
  • My Impression: I often use Claude Design / Code along with ChatGPT. I personally like the shapes in the second slide for the pilot to full rollout. It got 4 points because some parts overlapped. Design can be converted to PPT/Canva. It's a shame the output speed was the slowest. *By the way, Claude Design asks good clarifying questions even if the prompt is poor, so it's useful when your prompt isn't fully set.

5) Genspark

うちた - inline image
うちた - inline image
  • Score: Design 4, Text 4, Editing 5, Speed 4
  • My Impression: About a year ago, I thought Genspark was the strongest, and it's still powerful. The editing is especially good. The AI editing is the most advanced, and it can export to both PPT and Google Slides. It also has plenty of templates for slide decks. I want to research this more.
うちた - inline image

6) Copilot

うちた - inline image
うちた - inline image
  • Score: Design 1, Text 1, Editing 2, Speed 2
  • My Impression: I generated images with Copilot, but the quality felt like it was from a previous era. Copilot users, please let me know if there's a better way! 🙇

Actually, for Copilot users, it seems better to prompt Copilot directly within PowerPoint. At a glance, it looks like Claude, so maybe Claude for PPT is running.

(Score: Design 4, Text 4, Editing 5, Speed 3)

It looked like this 👇

うちた - inline image
うちた - inline image

Comparison Results

うちた - inline image

I mainly use Claude and ChatGPT for slide creation and thought it would be a landslide victory for them. But actually, other tools seem quite usable too.

Two things became clear:

  1. Despite the same instructions, the visual results varied more than imagined.
  2. However, I felt that with the right prompt, you can create effective materials. For Copilot users, prompting within PPT is perfectly practical.

Closing

Thank you for reading.

I will continue to post specifically about the "business application" of AI.

  • AI x Documents
  • Tips for document creation
  • AI x Business Application └ Automated email replies, automated quote creation, automated Salesforce entry, automated meeting prep
  • AI used: Claude / ChatGPT / Gemini / Copilot, etc.

I will deliver not just how to learn AI, but how to use it for work. Thank you for your support."

Переделать в YouMind

Превратите одну вирусную статью в полноценный рабочий процесс создания контента

Собирайте источники, расшифровывайте паттерны, создавайте активы, пишите черновики и публикуйте контент из одного рабочего пространства ИИ.

Исследовать YouMind
Для авторов

Превратите ваш Markdown в аккуратную статью для 𝕏

Когда вы публикуете длинные тексты, изображения, таблицы и блоки кода, форматирование в 𝕏 становится мучением. YouMind превращает полный черновик в Markdown в чистую статью, готовую к публикации в 𝕏.

Попробовать Markdown для 𝕏

Другие паттерны для анализа

Недавние виральные статьи

Смотреть другие виральные статьи