Maya never 429s, captions stay in sync, exports run in the cloud
Multi-model failover for AI, synced caption generation across both panels, retry button on failed messages, and Cloud export as the new default.
Maya doesn't fail when a model rate-limits
The AI chat and batch tools now fail over automatically to a backup model if the primary is rate-limited. You'll see the typing dots pause briefly, then Maya continues from the next provider — no error, no lost message.
Retry a failed message inline
If a chat message does fail, it now shows a small "Not delivered — {reason}" row below your bubble with a Retry button. Click it and Maya re-runs from that point, with orphaned tool calls cleaned up so the stream doesn't choke.
Caption generation is synced
Starting captions from the Subtitles panel and from the AI chat now show the same countdown in both places, and swap to "Done in Xs" when it finishes. Generated subtitles are auto-chunked to 6 words max so they fit one line on mobile.
Split a subtitle in two
The Subtitles side panel now has a scissors icon on each caption. Click it to split at the midpoint word — word-level timings and style are preserved on both halves. Useful for breaking up long lines from Whisper.
Follow-up chips in the AI chat
After Maya finishes a task, you'll see a couple of suggestion chips below her reply — e.g. after generating captions: "Remove filler words", "Check if this is engaging", "Add background music". One click and she's onto the next thing.
Cloud export is now the default
Exports now run on our servers by default. That means no more Chrome "this tab is using a lot of resources" warnings, faster processing for longer videos, and more consistent quality. Pick "Browser" in the export dialog if you'd rather encode locally.