Issue Description

What : In Regular Capture Mode, when “Automatic Font Size” is enabled, the size of the first captured text element sets the font size baseline for the entire translation output.
When : This becomes a problem when the first detected element is a large-format title, as it causes all subsequent translations —regardless of actual size in the source— to appear disproportionately large.

Issue Demonstration

1️⃣ Capturing begins with a title element.
Result: all translated text is rendered in the oversized title font.
 

2️⃣ Capturing starts after the title.
Result: translation font sizes appear truer to context. Though still not perfect as in this example, the game uses very large fonts. The red outline indicating where the previous demonstration ended.

Ideal Solution

Ideally, the translation font sizes should reflect the relative size and hierarchy of the original elements, not just the first one encountered. I do not believe the aim of Regular Capture is to replicate the Automatic Capture mode’s accurate text positioning, but at least having relative sizes would help a lot.

Why this matters

• Enhances usability: by preventing large capture areas from going off screen.
• Helps readability: by preserving a semblance of UI integrity across translated content.

  Thank you for your consideration.
🙏

Heya,

This is an easy fix and something I've planned to incorporate to the manual mode too. Its purpose is indeed to translate more uniform texts rather than large chunks, but adding the sizing functionality that exists in automatic mode won't break anything.

Thank you for the suggestion!

Hiya mate,

This was added in 0.4.6. Hope it works much better for ya and thank you for the nudge! :)

Testing with version 0.4.6

The capture with the title element is indeed now a bit more readable.
 
The underlying issue of better OCR is still to address, but this is a step in the right direction. 👍

And here is a capture with the default validated Noto font to give it a fair shake.