Scan2Speech — photos of book pages → audio

Add your OpenAI key

Stored only in this browser, sent directly to api.openai.com and nowhere else. Get one at platform.openai.com/api-keys.

Advanced settings (models & voice)

Vision model Any vision-capable chat model. gpt-5.5 (default) tested most faithful on dense non-English pages; gpt-4o/gpt-4.1 are good fallbacks. Avoid -mini (misreads). Page language (optional) A hint that sharpens transcription accuracy. TTS model Voice Max characters per audio chunk Kept under the speech endpoint limit. Text is split on sentence boundaries. Voice instructions (gpt-4o-mini-tts only) Auto-straighten & crop new photos (orient, deskew, trim to text)

Add page photos

🖼️

…or choose existing photos (or drag them here)

They start transcribing automatically and are auto-straightened & cropped. JPEG / PNG / WebP (HEIC where the browser supports it). Pages are ordered by when each photo was taken (oldest first) — you can reorder them below.

Add your OpenAI key

Add page photos

Review & listen