Sigma-blade kneaders keep the trough on Wan 2.7 I2V at 15cr

Wan 2.7 Image to Video is 2–15s, 1080p, 15cr; audio is null, so do not claim native sound. Animate the photographed sigma blades; a generated trough with a different volume is a different mixer.

A sigma-blade listing that books is the iron you already cast: two Z-blades in this saddle, this drive side, this working volume. Text-to-video “industrial dough mixer, golden hour, shop VO” will invent a cousin trough and a different litre count. Photograph the mixer. Move the blades. Leave the bowl alone.

The trough photo is the contract

Photograph the kneader you sell. Square to the blades if the cover is off, or three-quarter with the trough readable if the cover stays on. Drive housing in frame. Discharge end in frame. Working volume is a number on a cut sheet, not a prompt adjective. If the still already shows a stamped litre mark, keep it. Do not prompt it off and hope Wan rebuilds the bowl.

Text-to-video will sample another guess of a mixer. That guess is a capability claim you did not photograph. Lock the still before you spend on motion is the routing rule: identity is cheap as a still and expensive as a video retry. Re-describing “300-litre sigma mixer” in the motion paragraph is how the model starts drawing a second trough.

Image-to-video starts from the frame you upload. Frame one is the file. Prompt camera and physics only. Blades turn. Dough folds. Camera locked. No new volume. No new feet. No extra discharge. If the still would fail a distributor audit — wrong blade count, a different gearbox, a lid you do not stock — stop. Reshoot. Motion will not save a cousin bowl.

Wan 2.7 I2V is 2–15s, 1080p, 15 credits, silent catalog

Wan 2.7 Image to Video is the identity-hold row: a still required, 2s–15s in one-second steps, 1080p ceiling, 15 credits. Qualities listed: 720p and 1080p. Three modes: first-frame to video, first-and-last-frame interpolation, or video continuation. Catalog audio is unset. Optional driving audio is a track you attach, not a soundtrack the model always invents. Do not treat this as a native-audio talking row. Quoted millwright lines do not become a voice.

First-frame: your upload is 0:00; prompt the fold, not a new shop. First-and-last-frame: both stills must be the same mixer — lid open and lid closed, or empty trough and a batch you photographed in that bowl. Two different volumes is a morph. Video continuation extends a take that is already on-model; it is not how you invent the kneader.

Two seconds is the honest test. The spec will let you go to 15. Identity drift likes the long end: extra blades, a taller trough, a drive that swapped sides. Until the bowl holds, stay on 2s, 3s, or 4s. 15 credits is the catalog figure, not a reason to skip the short test. Do not write a shop bed into the Wan prompt and wait for sound. Score later. The file is silent unless you attached a driving track, and even then the catalog does not promise a native mix.

A generated trough is a different mixer

Working volume, blade geometry, and drive side are claims. A buyer matching iron will notice a 200-litre cousin before they notice your logo. Do not morph trough volume. If the still shows a 150-litre saddle, the clip may not finish as a 300-litre body. If the still shows two sigma blades, the clip may not grow a third. Those are SKUs.

Do not ask Wan to typeset litres onto the bowl in flight. Video models redraw letterforms every frame. Add a text overlay burns one authored line, whole clip, top / center / bottom. That is the working volume, the blade type, a CTA. It does not transcribe. A mute mixer has nothing to listen to. Overlay after picture lock. Swap 150 L sigma for the next row without regenerating the trough.

If the photographed stamp is already sharp and already true, keep it. Do not “enhance” the litres. If the mark is unreadable, overlay the string from the cut sheet. Do not generate a fuller trough so the volume “looks used.” A constructed millwright walking “what you should knead in this bowl” is a talking head you do not need. Keep lips off the mixer.

Product video holds the object; it does not recast the bowl

AI product video is the object-hold door when you have two or three clean angles of this kneader — trough, drive, discharge — and you want the camera to move around the iron you already shot. Reference photos are the contract. Prompt the scene, not a second mixer. Zoom to 100% before you keep a take. Fine type on a nameplate is the first thing a generator loses.

Use product video for an orbit of the photographed machine. Use Wan 2.7 I2V when the blades must travel on one approved still. Do not mix the jobs into one paragraph that also asks for a shop VO. Pause on the saddle. If the freeze-frame would not pass the catalog, the video will not pass the pallet. Animate the photographed sigma blades. Keep the trough.

FAQ

Can I text-to-video “sigma mixer, 300 litres” and skip the photograph?

You can. You will recast the trough and grow a volume you do not ship. Kneader identity is the still. Wan 2.7 Image to Video requires that image. 2s–15s, 1080p, 15 credits, catalog audio unset.

Does Wan 2.7 add a shop soundtrack if I ask for one?

No. Catalog audio is null. Optional driving audio is a track you attach. Do not claim native sound. Score later. Overlay litres. Do not file a bug on a row that does not list a mix.

When is first-and-last-frame legal on a kneader?

When both stills are the same mixer. Lid open to lid closed. Empty trough to a batch you photographed in that bowl. Two different volumes is a morph. Do not interpolate a 150-litre saddle into a 400-litre cousin.

Why not a talking millwright over the blades?

Because blade travel is visible on the photographed iron, and working volume is a number. A constructed tech is a second file. Hold the trough. Caption a human only if a real human is speaking.