Wan 2.7 Reference to Video: the stack is the brief (2s, 3s, 4s, 1080p, 18cr)
Wan 2.7 Reference to Video is reference-to-video. Upload the stills that must appear. Do not spend the whole budget because the slider goes that far.
Wan 2.7 Reference to Video is reference-to-video. Upload the stills that must appear. Do not spend the whole budget because the slider goes that far. The catalog description: “generate videos from up to 5 reference images and/or videos, with optional first frame and voice timbre control.” That mix is the brief. Five stills, five clips, an optional opening frame, a timbre — not a 10-second slider maxed out on a stack you have not frozen.
Wan 2.7 Reference to Video requires an image. It also accepts video on reference_video, up to five video inputs. Max output is 1080p (720p is on the quality list). Aspect ratios: 16:9, 9:16, 1:1, 4:3, 3:4. Durations: 2s through 10s. Credits: 18. Features: reference_to_video, multi_reference, first_frame, voice_clone. Native audio is not a boolean on this record; timbre control is.
Five images and/or five videos — not “more identity”
Happy Horse 1.0’s reference row takes nine stills and character placeholders. This Wan row takes five images and/or five videos, plus an optional first frame. That is a different contract. A reference video is motion you want repeated or a look you want held, not a second chance at a face. If the five stills already disagree about the jawline, adding a sixth file (you cannot) will not help, and adding a reference clip of a different person will make it worse.
Reusable characters and products is how the five stills stay the same five stills next Tuesday. Optional first frame is how you pin the opening composition without turning the job into a first-last interpolate. First-last is a two-still travel. This row is a cast.
Voice timbre control is not a new actor. It is a way to keep a voice colour across takes. If you need a cloned identity as a pipeline, the longer write-up is Wan 2.7 reference-to-video with voice clone. Do not confuse timbre with voice cloning as a separate engine. On this row, timbre is a control on the generate.
2s is a stack test. 10s is a scene you already trust.
The duration list is 2s, 3s, 4s, 5s, 6s, 7s, 8s, 9s, 10s. Two seconds is short enough to see whether character and product survived. Four seconds is a full social beat. Ten seconds is the ceiling. The slider going to 10s is not a discount on identity risk. Drift at 2s becomes a 10-second cousin.
1080p is the ceiling. There is no 4K token on this record. If the delivery is 4K theatrical, this is not the generate. For Reels and PDP loops, 1080p is the ship file. 720p is available when you are still proving the stack.
18 credits is the sticker. It is not 28 (Happy Horse 1.0 reference) and it is not 40 (several 1080p text-to-video rows). Different input shape, different bill. Do not pick Wan 2.7 reference because it is cheaper than a 40-credit scene row if you do not have the five files. You will spend the 18 credits inventing a cast you meant to upload.
First frame and timbre are optional. The stack is not.
Requires_image is true. A prompt with no stills is a different Wan row — Wan 2.7 text-to-video if the job is a scene from words. Optional first frame means you may pin the open. Optional timbre means you may steer the voice colour. Neither optional control replaces the stills that must appear.
The Wan provider roster is the family. The category ranking is best reference-to-video. Use this page when the job is Wan 2.7’s mix: up to five images, up to five videos, 2s–10s, 1080p, 18 credits, first frame optional, timbre optional. Upload the cast. Generate 2s. Then buy 8s. Do not buy 10s because the control went that far.
FAQ
Can I skip stills and only upload reference videos?
The record requires an image and accepts video as additional reference (max five). Plan on at least one still. Video-only is not the shape this row advertises.
How is this different from Wan 2.7 Pro Edit?
Pro Edit is an image editor: 8 credits, up to four references, stills out. This row is video: 18 credits, up to five images and/or videos, 2s–10s, 1080p. Do not edit a poster here.
Does 10s at 1080p always cost 18 credits?
18 is the catalog sticker for the row. Durations run 2s–10s. Prove the stack at 2s–4s. Lengthen when the action needs it, not when the slider tempts you.
Is native audio guaranteed?
Audio is not a true/false flag on this record. The description names voice timbre control. Write the sound you want. Inspect the take. Do not assume a multilingual lip-sync feature that belongs to a different family.