Maybe it's a caretaker bot outliving its family, or a factory unit that starts saving the birds that nest in the rafters. Describe your machine and what it wants, and you get a short film where the same chassis — every dent, every decal — appears in every shot.
Your robot's build is a casting decision: you approve its design from generated portraits, and that exact body — proportions, wear, markings — holds across the whole film.
Robots act with their bodies. The direction stages hesitation, tenderness, and menace through how the machine moves, so it can carry scenes with no dialogue at all.
The child, the engineer, the scrapyard dealer — supporting humans are cast with the same portrait approval and consistency as your machine.
Music matches the story's temperature — warm and sparse for a companion bot, industrial and cold for a rogue unit — and never smothers the sound design.
Dreamz is also a Claude connector (MCP) — describe your film to Claude and the finished video comes back in the conversation. If Claude is already your workspace, your film studio now lives there too.
Tell Dreamz about the robot story films you want — story, mood, who's in it. The director agent writes the script and casts characters with you, in chat.
Scenes generate live on your dashboard — visuals, voices, music and editing handled for you, with the exact price quoted before anything is charged.
Watch the finished cut, request changes in plain words, then download it — everything you make is commercially yours.
The robot short is one of the most attempted forms on the internet and one of the least often landed, and the reason is a mismatch between what the maker finds interesting and what the audience does. Makers love the design. Audiences love the want. A beautifully rendered chassis doing nothing has no story in it; a battered unit that keeps sweeping a floor nobody walks on anymore has a whole film. The genre's engine is a machine with a purpose it was given, meeting a world where that purpose no longer fits. That is why caretakers outliving their families, factory units adopting something, and decommissioned soldiers still standing watch keep working — the want is legible in thirty seconds without a word. Before you write, finish this sentence: my robot was built to do X, and it is still doing X even though Y. If you cannot fill in Y, you have a design, not a story. If you can, you barely need dialogue at all, and the film will be better for it.
Describe the build and the want in the same breath. A brief with enough teeth: "Two-and-a-half minutes, 16:9. UNIT-7 is a coal-yard sweeper in a shut-down power station outside Katowice — waist-high, treaded, one articulated arm, a sun-faded yellow shell with a hand-painted number on the flank and a magpie's nest wedged in its shoulder cavity. It has swept the same empty yard every day for six winters. In March the demolition crew arrives. It keeps sweeping around their boots. The foreman's kid chalks a face on it. Ends with the yard empty, swept, and the unit gone. No dialogue at all." The chat asks what is missing — aspect ratio if you had not said, whether snow is on the ground. It quotes a price in the chat; that quote is the exact charge, and nothing bills until you approve. Then it writes the screenplay and shows it to you before a frame renders. For a robot film, read the screenplay for physical action specifically: does the unit's behaviour stay consistent, does it ever suddenly become able to do something it could not do in scene one? Say "the arm can't reach above chest height — restage scene three" and it is fixed in text. The finished film arrives in the same chat, commercially yours.
Your design choices are casting choices and they determine what your robot can act with. A head that cannot tilt cannot show doubt. No eyes at all means every emotional beat has to be carried by the whole body — which is harder to write and often more moving. One light instead of a face gives you exactly one expressive channel, and one channel used well beats five used vaguely. So choose your expressive vocabulary deliberately, name it in the brief, and then write only beats that vocabulary can play. "It hesitates" is not stageable for a treaded box; "it stops, reverses six inches, then continues" is, and it means the same thing. The wear on the machine is doing narrative work too: dents are history, a repair in the wrong colour is somebody who cared, a factory-clean finish says this thing has no past. Because the design is approved from stills before production, spend that gate properly — check silhouette readability at a distance, and make sure the marks you are relying on for continuity are big enough to survive a wide shot.
The most common failure in the last thirty seconds of a robot short is a line of dialogue that explains the feeling the images already produced. If the machine says "I was built to serve, but I have learned to love," the film is over in the bad way — you have converted something the audience felt into something they now merely agree with. The genre's strength is that a machine cannot say what it means, so it must do something instead. Keep the resolution physical. It gives the thing away. It goes back to work. It stops. It stands in a spot with no functional reason to stand there. If your robot does speak — and it can, in a performed voice — the discipline is to make it speak only in the register of its function, and to spend real lines the way a western spends bullets. Two or three across a whole short is plenty. Write the ending as a stage direction with no words in it, read the screenplay to confirm nobody has helpfully added a summary line, and let the last shot be quiet.
Describe it — boxy or skeletal, chrome or rusted, one eye or none — and approve the generated design before production. The film is shot only after you sign off on the look.
Yes, with a native performed voice, or it can stay mute and act through movement and framing. Both approaches are staged deliberately in the script.
Prices are quoted exactly upfront in chat before anything is charged, and short videos start at a few dollars. You approve the number first.
No. The whole pipeline — script, robot design, scenes, voices, edit — runs from your chat description. There is no software to learn and nothing to upload.
Yes. Your approved robot design can return in future films, so a one-off short can grow into an episodic series with a consistent lead.
Yes, and facelessness often helps — the audience projects, and projection is stronger than what you could have shown them. What you must give it instead is consistent, readable body behaviour: a signature movement, a rhythm it keeps, a thing it does when it does not know what to do. Write those into the brief as physical rules rather than emotional labels, then check the screenplay stages them as actions rather than describing the robot 'feeling' something the camera cannot see.
Anchor recognition on two or three describable features that survive scale and lighting — a specific colour of shell, a number on the flank, an asymmetry like a missing panel — rather than on overall styling, which is where drift shows up first. Approve the design from stills, keep those same descriptors in every later brief, and reuse the approved design as your starting point. Continuity is easiest to keep when it depends on things you can say in words.
You can brief a general class of machine — a quadruped, an industrial arm, a warehouse mover — without invoking a specific company's product, and that is the safer and usually better creative path. Copying a real commercial design ties your film's look to someone else's brand and gives you a machine the audience already has opinions about. Describe the function and the physics you want and let the design be yours.
Yes. Dreamz ships an official Claude connector (MCP): connect it once in Claude, then describe the video you want in any conversation — pricing quotes, drafts and the finished film are delivered right in the chat. The website and the Claude connector share the same account and pricing.
One robot, described in chat, delivered as a directed film.