Tieback-anchor heads: Seedance 2.0 Fast R2V at 25cr, 720p, 4–15s

Seedance 2.0 Fast Reference to Video is 4–15s, 720p, 25cr discounted, native audio — reference the photographed head. Helical piers already spent MiniMax H3 torque at 26cr; this leftover is the tendon head, not another torque pass. /for/contractors is the crew page.

A tieback listing that reads is the head you tension: bearing plate, nut, strand or bar, the wall face behind it. Helix pitch disappearing into a pad is a different leftover. Write the head. Feed photographs of this head. Leave the mast to the helical-pier post.

The tendon head is leftover from the mast

The buyer already knows what a drilled wall looks like. They need to recognise the hardware you actually stock: plate diameter, strand count, the hex on the nut, whether the head is a bar or a multi-strand tendon. A paragraph that says “excavation tieback, cinematic soil nail, suburban retaining wall” will invent a cousin plate and a cleaner shotcrete. That is a different SKU.

Reference-to-video is the door when the photographed head has to survive a restage — a new cut, a different light, a 9:16 crop of a bench still you already own. You are not asking the model to guess a head from words. You are handing it up to nine stills of the real object (and, if you have them, short video and audio refs) and asking it to hold that identity while the camera moves.

Image-to-video is a different contract. That tool treats an upload as frame one: same excavation, same camera, motion on the plate you already hung. If the brief is “this photo, now a slow push,” stay on I2V. If the brief is “this head, in a wall we did not shoot,” you are on reference-to-video. Mixing the two is how a hung still gets restaged into a cousin wall.

Prompt camera and light only. Slow orbit that keeps the nut readable. Hold on the plate. No new strand count. No new house above the cut. If the still includes a stamp or a serial, keep it. Do not prompt it off and hope the model rebuilds the hex.

Seedance 2.0 Fast R2V is 4–15s, 720p, 25cr discounted, native audio

Seedance 2.0 Fast Reference to Video is the catalog row for this leftover. ByteDance. Category is reference-to-video only. Catalog sticker is 25 credits, marked discounted. Durations: 4s through 15s, every integer second. Max output 720p. Qualities listed: 480p and 720p. Native audio is on. requires_image is false — the row will run without a still — but a head you did not photograph is a guess you will then have to defend.

Reference slots: up to 9 images, 3 videos, 3 audio files. That is the stack. Front of plate, nut in profile, strand tails if they are yours. Do not pad the stack with a competitor’s catalog PDF. Aspects include 21:9, 16:9, 4:3, 1:1, 3:4, 9:16, plus auto. Pick the ratio of the cut. Describing “vertical” in the prompt while leaving aspect on auto is how you get a 16:9 hero you then crop into a 9:16 hole.

Audio is native. Write the mix you will stand behind — a quiet pad, a distant excavator — or write silence into the brief. A silent brief still gets a soundtrack you did not choose. Do not prompt quoted kips and expect a title card. Overlay after lock if a load string must appear.

Start at 4s or 5s until the hex holds. Fifteen seconds of a melting nut is a long miss. Padding a 5-second hold to 15 seconds is how the back half invents a second plate.

Helical piers already spent the torque pass

Helical piers are MiniMax H3 torque at 26cr already owns the install mast: drive head, helix pitch, 2K text-to-video, 5–15s, 26 credits per second, audio unset. That listing is torque you can see. This leftover is the tendon head. Do not rewrite the pier post as an excavation. Do not send H3 a paragraph about “tieback tensioning” and call the missing plate a technicality.

MiniMax H3 Text to Video has no still slot. It will sample another guess of a head. Average is not your nut. Average is a plate that does not match the hardware on the truck.

Keep the jobs apart:

Brief	Object	Row
Photographed head must survive a restage, 720p, 4–15s	Tendon head as identity	Seedance 2.0 Fast R2V
Hung still is frame one, same wall	Site plate	Image-to-video, not this R2V row
Torque mast from words, 2K, no still	Install as a scene	MiniMax H3, already spent

If a sentence contains “bearing plate,” “strand,” or “hex nut,” you are on this head. If it contains “helix,” “torque motor,” or “lead section,” you are on the pier post. Text-to-video of a head “because the wall should look retained” rebuilds a cousin plate in the middle of a mast shot.

Reference-to-video is the mechanic: the files pick the subject; the prompt picks the shot. A three-word “tieback cinematic” hands both jobs back to the model. Feed the head. Write camera, light, and duration. Then stop.

Contractors get a different /for page

/for/contractors is the crew page: process transparency, demo-day time-lapse, a problem-and-fix on site, a final walkthrough. Two posts a week. 9:16. Facebook and Reels. That page is not an OEM catalog of tendon heads. Name the difference. Do not rewrite /for. Do not steal its hook for a plate orbit.

OEM content that books a specifier is the head, the plate, the strand count. Crew content that books a service call is a real cut, address stripped. Mixing them is how a catalog clip becomes a fake estimator walking “what you should do before the next rain.”

A photoreal estimator prescribing tension is an expert persona you do not need. Hold the head. Burn a load line after lock if it must appear. If a human must speak, that human is a real named person on a different day. Generation can still do the plate. It cannot be the licence.

The test: is the point a photographed head that has to survive a restage, in 4–15 seconds, at 720p, with native audio? Yes → Seedance 2.0 Fast R2V. If the point is a helix on a pad you will invent from words, you are on the pier post. If the point is this hung still as frame one, you are on image-to-video.

FAQ

Does Seedance 2.0 Fast R2V require a photo of the head?

No. requires_image is false. Run it without a still and you will get a cousin plate. If the photographed nut must survive, feed it as a reference — up to nine images — and describe only the camera.

Why 25cr discounted instead of MiniMax H3 at 26cr?

H3 is 2K text-to-video, 5–15s, 26 credits per second, audio unset, already spent on the torque mast. This row is reference-to-video, 720p, 4–15s, 25 credits discounted, native audio. Pay it when the head you photographed has to hold while the wall restages. A 2K invented tendon is a different listing.

Is this the same as uploading one still to image-to-video?

No. Image-to-video treats the file as frame one. Reference-to-video treats the files as identity. A hung still that must stay the hung still is I2V. A bench still that must appear on a wall you did not shoot is this R2V row.

Will Fast add a VO if I prompt “foreman explains tension”?

Native audio is on. That does not make a constructed estimator a spec. Write silence or a pad. Overlay the kip string after lock. A talking expert on this leftover is a second file, with a real named person, or no file.