I kept seeing garbled Arabic text in AI-generated videos, so I built a tool that fixes it
Every Arabic content creator I know hits the same wall with AI video tools: ask for on-screen text and it comes out as unreadable, disconnected shapes — because these models don't do real Arabic letter-shaping or right-to-left reordering.
I got tired of it and built Montaj AI: you write your idea (in Arabic or any language), it generates the scenes with AI, and instead of asking the AI model to draw text (which never works), I render captions separately with real Arabic font shaping and burn them onto the video with ffmpeg/libass — the same technique used for professional movie subtitles.
Just launched on Product Hunt: https://www.producthunt.com/posts/montaj-ai . One free video, no card needed. Would love feedback, especially from anyone who's fought this same battle with non-Latin scripts.
The workaround is more interesting than the generation itself. Separating text rendering from the video model seems like the kind of constraint that could make this useful beyond Arabic too.