Subtitles feel solved now — but how do you translate text inside videos?

It feels like speech and subtitles are mostly solved now.

But one part of video localization still feels surprisingly manual:
text that appears inside the video itself.

• Slides.

• UI screens.

• Signs.

• Product labels.

• Graphics with embedded text.

If you want to translate those, the workflow usually becomes messy — recreating graphics, editing frames, or just leaving the text as-is.

Curious how others deal with this today:

• How do you handle translating text that appears inside videos?

• Do you edit it manually, rebuild the visuals, or use some other workflow?