Planetariums: the dome is an architecture plate, not a talking astronomer

Hailuo silent I2V of the empty dome you photographed. Overlay show times. A constructed host explaining the sky is the wrong row.

A planetarium clip that informs is an architecture plate: the empty bowl, the projector pit, the seating rake, the door you actually operate. It is not a constructed astronomer walking the ecliptic, a generated sky that is not your projector, or a host telling a stranger "here is what you should know about Mars." Photograph the empty dome. Animate that file. Burn the showtimes you already publish. Leave the lecture off the generate.

Photograph the empty dome

The honest still is the room you run: seats in, console dark or generic, no head on a mic, no live sky that belongs to tonight's show. The public already knows what a dome looks like. They need this dome, then the hours string the box office already uses.

Do not send "photoreal planetarium, talking astronomer, photoreal Milky Way" to a text-to-video row and call it the venue. Text-to-video will decide a bowl. That bowl is not yours. You cannot overlay a show title onto a room the sampler built.

If the still has to be this theatre, photograph the empty house. Crop people. A generated visitor is a private person. Showtimes are type.

/for/museums is a different leftover: one object, one story, a curator handling an artefact. This page is the dome as architecture. Do not rewrite the museum /for as a talking astronomer. Prompt furniture and light only. Empty seats. Projector idle. If a face is in the brief, the brief is wrong.

Hailuo silent I2V holds the bowl

Hailuo 2.3 Standard is the motion row for this leftover: image-to-video and text-to-video, 6s or 10s, 1080p, 28 credits, audio false. Silent. That is the point. Quoted lecture lines in the prompt do not become a soundtrack and do not become a host.

Use the image-to-video path. The image-to-video tool is the door. Frame one is the photograph of the empty dome. Prompt the travel: a slow push down an aisle, a light rake across the ribs, camera locked, no new seating. Do not re-describe the architecture. Re-describing it is how the model paints a cousin of your venue.

There is no duration between 6 and 10. You do not get 8s. 6s is a short hold. 10s is for a gesture that lasts 10 seconds, not a second floor of the building. Do not "upgrade in the prompt." Text-to-video on Standard will invent a dome. Label that as a render, or photograph the house you operate.

Audio is false. There is no host VO on this row. Do not write a lecture into the motion prompt and expect a stem. Score later, or pick a native-audio row on a different day when a real named person is in frame.

Overlay show times; Paper survives a busy sky

Showtimes are a string the box office already uses: "Friday 7pm, The Sky Tonight." "Saturday matinee 2pm." Write that string. Do not ask Hailuo to letter it into the dome.

Add timed text overlays to a video burns lines you wrote at start_sec and end_sec. Nothing is transcribed. "This week," then the Friday title, then the door time is three authored beats on a locked plate. Font, size, colour, and position apply to the set, not per line. If you only have one persistent badge for the whole clip, that is add a text overlay: one string, whole duration, top / center / bottom.

Auto-captions listen. A mute dome has nothing to transcribe. Running a transcript pass hoping "show times" appears is a spend on silence.

A starfield and a ribbed bowl eat a one-pixel stroke. Paper is Montserrat black type on a white rounded card, kind=subtitle. Use that family when you later caption a real recorded show: speech on a busy plate, the card is the contrast. If nobody spoke, Paper is the wrong tool for silence. Showtimes on a mute architecture plate are overlays you authored. Do not regenerate the dome to "add a white box."

Crop to the ratio you will ship before you overlay. Picture lock, then type. Hours change. The empty dome should not.

A constructed host is the wrong row

A talking astronomer is a presenter job. Hailuo 2.3 Standard will not grow a safe lecture from a monologue in the motion prompt. A photoreal host explaining the sky is a different catalog row, and for this venue it is the wrong one: you do not need a constructed expert to show a bowl. You need the bowl, then the times as type.

Do not prompt:

• An astronomer on a mic under the ribs
• A child in a seat pointing at a generated sky
• Hands on a console with readable show files
• A "sky instructor" walking what Mars means tonight

Those are people. Architecture still, silent motion, authored type. If someone must speak, put a real named educator on camera on a different day — and then caption that take with Paper so the card survives the starfield. If the first three seconds look like a host, rewrite until they look like a room. Photograph the empty planetarium. Hailuo the travel. Overlay the showtimes. Skip the constructed astronomer.

FAQ

Can I text-to-video a generic dome if we do not have a photo yet?

You can, on Hailuo 2.3 Standard's T2V path, and you must label it as a render. It will not be this theatre. Get the empty-house photo. I2V from that file is the architecture job.

Does Hailuo 2.3 Standard include a host VO?

No. Catalog audio is false. 6s or 10s, 1080p, 28 credits, silent. Overlay showtimes. Score later. A talking astronomer is a different row and the wrong leftover.

One overlay or timed lines for show times?

One persistent hours badge is a single overlay on the whole file. "This week," then Friday, then the door time is timed overlays, with timestamps you set. Neither job transcribes. Neither job belongs in the Hailuo prompt.

Can we generate an astronomer if we disclose it is AI?

Disclosure does not make a constructed host the architecture plate. Keep the empty dome. Keep the overlay. If someone must speak, that someone is a real named person, not a generate.