Short answerAfter two paid attempts on the same shot without an accepted result it must stop, say what it tried and offer alternatives, so it never burns credits in a loop. Any reply from you unlocks one more attempt. Yes and No answer its last question; Confirm plan approves a plan card.

Every tool that spends credits on a shot (a retake, a shot re-render, a group re-render, a split, an insert) counts against that exact target: this episode, this segment, this group, this shot. After two attempts without a take being accepted, the next attempt is refused by the system, not by the agent's mood. It then has to:
If the shot is short and static, it may run the free still-shot route straight away instead of asking.
Your next message. Each reply grants one more attempt on that target, so "try once more with the reference attached" works, and so does "leave it". An accepted result resets the counter to zero.
The rule exists because a re-render of the same text rolls the same dice. Two misses usually mean the text or the references need changing, not the dice.
The ✕ next to the input cancels before the next step. Renders already kicked off still finish and report back. To stop a whole episode run, ask it to cancel the run; queued renders finish on their own.
"Something failed on my side (…)" is a tool error; retry once. If it declines to publish, approve or quote a price, that is by design, see what it can do.
Still stuck? Email support@dawnwood.ai with your account email, the series, the episode and the segment or group — see what to include.