GPT-5 Support and Direct PDF Reading for Significant Improvement of Speech Script Generation Quality

We have significantly updated SpeechSlide AI, our automatic presentation video generation service, to support the latest large language model "GPT-5" and its lightweight, fast version "GPT-5 mini".

Previously, scripts were generated based only on text extracted from PDFs. With this update, the AI can now read the PDF file itself directly, enabling more natural speech script generation that takes into account slide layout structure, element placement, and the relationships between figures and tables. This is especially helpful for figure-heavy materials such as academic and research presentations.

We have also completely redesigned the prompts used for speech script generation, optimizing sentence flow, emphasis points, and natural spoken-language tone, so the resulting scripts sound smooth when read aloud by TTS (Text-to-Speech).

Generated scripts can be freely adjusted in the editor. Please try the improved script generation.