PJ Ace reports a $28,775 budget for the latest Nexus film

• The latest Nexus film cost $28,775, according to PJaccetturo's announcement, with $5,000 to $7,000 attributed to animation.
• Production involved 2.5 weeks of full-time director/animator/editor work, as PJaccetturo described it.
• Director Chris Capel separated framing, staging and performance into different passes, preserving successful shots through techniques detailed in the production notes.

Soldiers running the wrong way were painted out in flat white, then regenerated inside the original frame. A ten-second close-up prompt assigns separate jobs to six image references and a voice-reference video, then specifies five hard-cut shots down to their dialogue timing.

PJ Ace reported the budget alongside 62,479,089 cumulative views across the Nexus series in his production thread. The new 3-minute-27-second episode features a red-skinned rebel, explosive-pustule artillery and goggled minions called Bobos, a gloriously gross assignment.

Ace and writing partner Dan Weller brought in director Chris Capel, whose revised, more photorealistic Bobo designs won them over.

The latest Nexus film

Ace supplied two additional production figures in replies:

• Animation: $5,000 to $7,000, according to his budget reply, roughly 17% to 24% of the reported total.
• Schedule: 2.5 weeks full time, which he described as director/animator/editor work.

Asked about the cost, Ace said the team already uses agents and works with $1,000-a-day directors supported by a full team.

Nexus is a Dreamina original series, and Ace thanked Dreamina for partnering on it. A reader asked whether an independently funded episode would earn its budget back; the thread supplies no revenue figure or partnership payment terms.

Capel's behind-the-scenes notes describe four techniques for keeping successful work while retrying the part that failed:

1. White paint-outs. When soldiers had the wrong scale and direction, Capel painted them out of a keeper frame in solid white. He reused that image as the start frame, supplied soldier appearance references and specified the replacement action. The prompt explicitly identified the white areas as placeholders that should never appear in the output.
2. Maya grayboxes. Simple shapes established character placement, scale and camera position before generation. A rendered still supplied composition; a rendered camera move supplied motion. Capel labeled these references as staging and camera guidance only, with finished appearances supplied separately.
3. Coverage before acting. Capel repeatedly generated a whole dialogue scene until the angles and shot order worked, ignoring performance during that pass. He harvested a clean frame from each shot, upscaled it and used it to regenerate that shot individually until the acting landed.
4. Physical, timed direction. Camera shake became an operator walking and being bumped by soldiers. Performance beats received explicit time windows, while successful takes could return as performance and voice references. For one deadpan officer reaction, an early nod read as sarcastic; describing a single flat head movement with a blank face produced the intended joke.

Capel specified optics, action, sound and continuity for each cut in the published scene prompts:

• The cheering rider: Ten seconds, five two-second close-ups. A separate face-detail reference was explicitly excluded from controlling composition, pose or camera angle. Dialogue was limited to “Yes,” “More” and “Release,” with wind, boots and armor but no music.
• The battlefield duel: Four shots over 6.5 seconds. In the final shot, the Imperial knight's fury becomes flat annoyance between 0.5 and 1.1 seconds, before explosions engulf both fighters.
• The puppet gag: Three shots over 5.8 seconds. The wonderfully fussy continuity rules keep the white doll on screen-left and the red-haired doll on screen-right, distinguishing screen position from the actor's hands. A reference showing the Bobo crouched behind a rock defines his ending pose, rather than his starting position.

The notes also report that shorter prompts with one clear job per shot outperformed longer ones, despite the elaborate multi-shot examples shared in the thread.

The production notes cite a 12-file reference ceiling. Dreamina's current reference FAQ assigns that limit to Seedance 2.0 and lists up to 50 multimodal inputs for Seedance 2.5, subject to the selected mode's accepted inputs.

The Seedance 2.5 product page also advertises 30-second generations, consistent with the notes' description of using 20-to-30-second coverage passes.