Short answers. Where something is not supported yet, it says so.
You pick 30, 45 or 60 seconds. The finished video follows the length of the generated voiceover instead, so the narration is never cut off mid-sentence.
A 720×1280 MP4, H.264 video with AAC audio. Captions are burned into the picture rather than being a separate subtitle file.
English only. Both the voiceover and the caption timing are tuned for it.
Yes. There is no watermark and no claim over what you make. What you are responsible for is having the right to post the story text you pasted.
Google, GitHub or an email address. No separate account to create.
Yes. Finished videos and their job records are removed from the server an hour after creation. Download yours before you close the tab.
Segments are picked at random from the full-length footage, and the picker avoids the timestamp it used most recently, so consecutive videos do not open on the same moment.
No. Caption timing is estimated from word length rather than measured from the audio, so on a long script it can drift by a fraction of a second. It is close enough for reading, not frame-accurate.
Most often the voice service was unreachable, or the server ran out of disk space. Failed jobs are retried twice before being marked as failed, and the error is shown on the job.
Not from the browser. The games currently bundled are the only footage available.