Ham radio Field Day: diegetic rig on a locked antenna, not a talking operator
Vidu Q3 from the mast still for key clicks. TTS the band plan as a voice file. A photoreal operator is a likeness, not a QSO.
Vidu Q3 from the mast still for key clicks. TTS the band plan as a voice file. A photoreal operator is a likeness, not a QSO.
A Field Day clip that books a visitor is the antenna you already photographed: mast, guy, feedpoint, the tent fly if it is in the plate. The sound that belongs on that clip is the rig — key clicks, wind in the guys. A constructed ham walking a contact is a person you did not clear.
The mast still is the Field Day plate
Photograph the antenna you actually raised: same tower or mast, same guys, same feedline run. Text-to-video will invent a cleaner park with a cousin yagi and a callsign you do not hold. That is a failed site, not a cinematic upgrade.
AI image to video keeps frame one as the file you sent. Labels on the box, the generator in the grass, the cracked picnic table behind the mast survive because they were already in the picture. The prompt is for physics — a slow push on the boom, a guy that does not grow a new element, light across aluminum that is already in the photograph. Re-describing the park in the prompt is how the model starts drawing a second field.
If the still includes a face, crop it. You do not need a person for this ad. You need the mast. A recognisable private backyard you do not have rights to cannot stay. Minors in the background are a hard stop. Do not prompt “ham radio Field Day, operators laughing around a picnic table, golden hour.” That sentence restages a gathering. Empty mast. Diegetic rig. Band plan as a file you attach.
Vidu Q3 from the mast still for key clicks
Vidu Q3 Image to Video is the still-to-motion row with native audio: 5s / 10s / 15s, 1080p, 16 credits, discounted. Audio is on. Write the mix you want — key clicks, a quiet generator, wind in the guys — or you will get a soundtrack you did not choose. A silent brief still gets sound. Prompt the rig, not a lecture.
Iterate at 5s. Fifteen seconds of the same boom is a long miss unless the shot actually travels. Aspects follow the still. Diegetic means the sound is in the picture: key, fan, guy wire. It is not a host. It is not a net control. Quoted VO in the Vidu prompt is how you get a mouth you did not cast. Keep lips out of frame.
If you needed a last frame — mast down to mast up — that is a first-last job on a row that accepts two stills. This Vidu slug is one still, then motion. Photograph the state you want to start from.
TTS the band plan as a voice file
The band plan is a string you already publish: frequencies, modes, the hours the club posted. It is not a QSO. Gemini 3.1 Flash TTS is a 4-credit text-to-speech row: thirty voices, natural-language style control, inline tags ([sigh], [whispering]), multilingual synthesis, multi-speaker dialogue listed. That is a voice file. It is not mast physics.
Write the band plan. Generate the read. Keep it third-person and posted: the hours, the bands, the park name. Do not write “here is what you should call” as if the voice were net control. Do not seed a club member’s voice. Preset voices only.
Add a voiceover to a video is the attach job: plate locked, voice file in, mix mode if you want the Vidu key clicks underneath. Two files, two meters. Do not ask Vidu to be both the clicks and the band plan. Native audio on the mast is the rig. The band plan is a bed you attach.
The mouth does not have to be in frame for a voice to become a presenter. A closed-mouth “work 20 meters now” under a mast still is still a host. A band-plan file is a schedule. Keep the grammar on the schedule.
A photoreal operator is a likeness, not a QSO
A generated ham at the picnic table is a private figure. A deepfake of a private person is not a style. “Aesthetic Field Day” is not a defence. Seeding a club headshot “for consistency” is a replica. Labelling the face AI does not create a grant. A QSO needs two people. You do not have either of them.
Do not generate:
- An operator with a headset and a log
- A child at a GOTA station
- A “typical ham” whose face is a blend of people you know
- Lipsync on a constructed mouth reading the band plan
Those are likenesses. Field Day content that invents a body is not a mast still. If a real named operator must speak, that is footage of that person on a different day, with a grant. Generation can still do the boom. It cannot be the QSO.
Club logos and other callsigns in the plate are other people’s marks. Crop them unless you have the right to show them. Your call on overlay type you wrote is a string. A model-invented callsign on a hat is a fabricated identity. The test is a pause. Mast matches the field: ship. Invented operator: you asked a talking row to be the QSO. Leave the picnic table empty.
FAQ
Can I generate a ham at the table if I crop the face?
No. A photoreal body at a Field Day station is still a generated person. Photograph the empty mast. Do not invent an operator so the antenna has a scene.
Does Vidu Q3 image-to-video invent the antenna if I only write a paragraph?
No — this slug requires the still. Frame one is the file. Text-to-video is a different row and will invent a park. Upload the mast you actually raised.
Why TTS the band plan instead of prompting Vidu to talk?
Vidu’s native audio is the rig mix: key clicks, wind, generator. Quoted VO is how you get a mouth. Gemini 3.1 Flash TTS is a 4-credit voice file of the schedule you wrote. Attach it with add a voiceover. Keep the grammar on the band plan, not on a QSO.
Is a talking AI operator acceptable if we disclose it?
Disclosure does not turn a constructed ham into a cleared likeness. Keep the mast. Keep the clicks. Attach the band-plan file. If someone must speak a contact, that someone is a real named operator on a different shoot, not a generate.