It has been a chaotic week. Got hacked three times. Someone hit our SES and sent lakhs of phishing emails. Then our server got compromised, and the site went down.
But we're back up, and we still shipped some updates to our text-to-video tool (outputs React/TSX instead of video files).
We built it for edtech content like code snippets, diagrams, and tutorials. Where clean text rendering matters.
Watching users, we noticed a mismatch. We built this with voiceovers in mind (as our course content was in text format). You write a script, we generate visuals to match.
But users would put in visual descriptions ("a cowboy sitting in a cyberpunk bar"), questions ("teach me about rice harvesting"), or anything except an actual script. That's what most other tools accept.
So now the system generates a script from your input first. You can review/edit it before the generation starts. Also added multiple voice options and a gallery showing what's possible.
Try it: https://outscal.com/
Happy to talk about the UX changes, and would love some honest and brutal feedback