Wan 2.2 Move is motion, not speech

Wan 2.2 14B Move is 8 credits, 720p, category motion-control. It copies the movement in a driving clip onto a subject in a still. It is not speech. It is not Wan 2.2 Speech to Video, which is 50-credit 720p lipsync from a still and a track. If you wanted a mouth to say a line, you wanted Speech. If you wanted a body to do a move that already exists on a file, you wanted Move.

This page is the motion-control row. It is not the talking-head mix-up, and it is not Kling’s 17-credit 1080p drive.

Motion-control is two files, not a sentence

The published card: content type video, category motion-control, image required, 720p, 8 credits, audio null, no duration chips. Description: advanced video motion control with 14B parameters. Length of the output follows the drive. Choreography lives in the video file, not in the text box.

File A — the still. Who or what should perform. Full body if the drive is full body. A headshot against a dance drive is how you get invented legs. Lock this first at image prices.

File B — the driving clip. Someone, or a product on a turntable, already doing the motion: walk, unscrew, orbit, a gesture you can mute and still read. Subject fully in frame. You have the rights.

If File B is missing, stop. Do not type she spins twice and points at the label. That sentence is image-to-video. Run I2V. Move will not parse it into choreography.

If File B is a 9:16 dance and File A is a 16:9 pack on a table, stop. You are about to transfer a full-body performance onto a bottle crop. Recrop or regenerate the still to the drive’s framing, or pick a drive that is a tabletop turn.

Confirm you are in the right mode:

1. Can you mute the drive and still understand the motion? If no, the pose is not readable.
2. Can you freeze the still and cast it in that blocking? If no, fix the still.
3. Do you need the mouth to say a line? If yes, this row will not help. Audio is null. Lipsync is a later pass, and motion transfer is body-level anyway.
4. Is the motion something you could describe in one verb and accept as approximate? Then I2V is cheaper. Motion control is for when the timing must be that file.

Motion control is the glossary object. Move is the Wan 2.2 SKU for it. Kling Video V3 Pro Motion Control is the other published drive on this catalog: still plus driving clip, 1080p, 17 credits, audio false. Same job family, different ceiling and sticker. Do not open Move because the prompt said “Kling dance” and 8 looked cheaper than 17 if you needed 1080p.

Speech is the other Wan 2.2 row

The mix-up this page exists for is the family name. Wan 2.2 ships more than one card. People hear “2.2” and paste a VO onto Move, or paste a dance drive onto Speech.

Wan 2.2 Speech to Video is lipsync: audio-to and image-to, image required, audio true, 720p, 50 credits. Still plus track in. Talking picture out. Facial expressions, body movement, and camera work in the description are performance on that plate, not a location scout. Wan 2.2 speech is not a talking head is that argument. This page will not restack it.

Move does not take a speech file as the product. Audio is null. A wav attached because you hoped the arms would hit the downbeat of a sentence is not how the card bills. The drive is a video of motion. The still is the cast. The output is 720p motion. There is no viseme pass.

If you needed both — a transferred walk and a line — you run two rows, in order:

1. Move (or Kling motion-control) for the body, 8 credits at 720p on this SKU.
2. A video-to lipsync later if a mouth now exists and a track must match it: Kling Lipsync at 7 credits, 4K, or VEED Lipsync at 4. Or Speech at 50 if you still only have a still plus a wav and you never needed a drive.

Do not write the line into Move’s prompt. Do not write the choreography into Speech’s prompt. Each card has one second file.

Wan 2.7 text-to-video is a different object again: 2–15s, 1080p, 15 credits, audio null on this catalog. That row invents a shot from a sentence. It is not Move. Sending “founder walks into frame and says the offer” to 2.2 Move is how you discover requires_image — or how you upload a random still so the job starts, then watch the wrong body do the right dance.

720p, 8 credits, no soundtrack

720p is the ceiling. There is no 1080p or 4K on this output list. That is a motion-transfer spec, not a cinematic master spec. If the still is a 4K pack you will inspect for type, this is the wrong ceiling. Kling V3 Pro Motion Control lists 1080p at 17 credits. Upscaling a 720p transfer and calling it a 4K product orbit is a round trip that throws away the reason you had 4K.

Eight credits is the listed sticker, not a USD price. It is also not a reason to drive a still you have not locked. Transfer on a plate you will replace is 8 credits of noise. Lock the still. Trim the drive to the motion you mean. A 30-second take with 20 seconds of waiting will transfer the wait.

Audio is null. Move will not write a score, a VO, or room tone as a catalog promise. Plan the mix after. Add music to video if the cut needs a bed. Native audio is a generate feature on Veo, Seedance, Kling O3 Pro. Move is not those rows.

Rights on the drive: trend rips from For You are not a performance you can ship. Shoot it, license it, or use a clip you own. Phone, locked off or stable, full body, even light, 3–10 seconds of the actual gesture. Clap or count if you need hits. Do not hide the joints.

Then:

1. Open Wan 2.2 14B Move (8 credits).
2. Attach the still. Attach the driving video. Both.
3. Prompt, if the field is there, is wardrobe, backdrop, mood — not choreography. Do not argue with the file in prose.
4. Generate. Watch at 1×. The beat should be the drive’s beat. If it is not, the drive was unreadable or the still could not be cast.
5. Do not add speech here. Score, VO, or lipsync after.

The AI video generator will not fetch a dance because you named a song. You attach the clip.

What Move is not

Refuse these inputs:

• A YouTube link in the prompt. There is no file.
• A still of a pose. That is a still. Motion control needs time.
• A native-audio generate you liked for the song. You can try using it as a drive if you own it and the pose is readable; you are transferring motion, not the song. Audio is null on the output.
• “Make it like last week’s Reel.” Last week’s Reel is only a drive if you import the file and have the right to use the performance.
• A VO and a headshot, because the slug says 2.2. That is Speech, 50 credits, still plus track.

If what you actually wanted was a product orbit: film a 5-second orbit of any stand-in, or use I2V with “slow orbit, label holds.” Do not attach a hip-hop drive to a serum still.

If what you actually wanted was a talking presenter: lipsync. Move will not lip-read a VO. Most transfer is body-level; expression and lips usually need a separate pass.

If you need a seat to attach the still and the drive, sign up. Bring two files. Do not bring only the word “Wan.”

FAQ

Is Move the cheap version of Wan 2.2 Speech?

No. Move is motion-control at 8 credits, 720p. Speech is lipsync at 50 credits, 720p. Different second file: a driving clip versus a voice track. Eight is not a discount on fifty.

Why would I pay Kling 17 credits if Move is 8?

Kling V3 Pro Motion Control lists 1080p. Move lists 720p. Same job family. Pick the ceiling the delivery needs. Do not pay 17 for a 720p feed file, and do not pay 8 into a 1080p master.

Can I make them talk on this row?

Not as a catalog promise. Audio is null. Transfer the body, then run a lipsync row if a mouth exists and a track must match it.