ceeon/aippt2
Overview
This skill converts article text into a polished, image-rich PowerPoint using a guided workflow. It produces an ASCII-based slide framework, applies a chosen visual style, generates AI images, and outputs a complete PPT package. The process is repeatable and optimized for consistent slide aesthetics.
How this skill works
The skill extracts core information from an article (type, audience, key points) and maps it to a slide structure rendered as an ASCII PPT framework. You pick a visual style and the system builds image prompts that combine style, page content, and text placement. Images are generated via a fixed image-generation API and then assembled into a final PPT.
When to use it
- Turn long-form articles into presentation decks quickly.
- Create style-consistent slides across many pages or chapters.
- Prototype visual concepts and layouts before manual design.
- Generate image-led slides when photography or illustrations are needed.
- Batch-produce slides for workshops, training, or client briefings.
Best practices
- Extract and prioritize 3–5 core points per slide to keep designs focused.
- Choose a single visual style per deck to maintain cohesion.
- Iterate image generation when text rendering has variability; expect multiple attempts for perfect Chinese text.
- Use parallel API calls for large batches to speed up image generation.
- Review and tweak ASCII layout before image generation to avoid rework.
Example use cases
- Academic article to lecture slides with diagrams and concept visuals.
- Marketing blog post to a product pitch deck with thematic imagery.
- Technical documentation to training slides with annotated screenshots and callouts.
- Executive summary to investor presentation with stylized cover and key-metric slides.
- Workshop guide to step-by-step instructional slides with consistent visual language.
FAQ
The workflow uses the gemini-3-pro-image-preview-4k model at https://ismaque.org/v1/images/generations (these parameters are fixed).
Can I change the generation API settings?
No. The image-generation endpoint and model are fixed by the workflow and must not be modified.
What if generated text in images is inconsistent?
Text rendering can vary; generate multiple variants and select the best. Adjust prompts to emphasize legibility and text placement.