Guides

    Velodromes lock the banking; a generated peloton is a field you did not enter

    Wan 2.7 I2V holds the painted apron. Extend only after the same empty banking holds. Text-to-video draws riders who are not on the start list.

    Versely Team6 min read

    Wan 2.7 I2V holds the painted apron. Extend only after the same empty banking holds. Text-to-video draws riders who are not on the start list.

    A velodrome clip that books a session or a hire is this bowl: painted apron, blue band, the banking a neighbour already walks past, the infield you actually mow. It is not a constructed peloton, and it is not a start list you did not enter. Photograph the empty track. Animate the paint. Leave the riders off the generate.

    Wan 2.7 holds the painted apron

    The buyer already knows what a bike looks like. They need to recognise this venue: apron colour, distance marks, the roof line of this cover, the infield furniture that is actually there. Text-to-video will invent a cousin bowl — a prettier banking, a different blue, a grandstand that is not on your lot.

    Image-to-video starts from the still you upload. Prompt camera and physics: a slow push down the apron, light on the paint, no one entering frame. Re-describing “packed madison at golden hour” is how the model starts drawing a field.

    Wan 2.7 Image to Video is the identity-hold row: a still required, 2s–15s in one-second steps, 1080p ceiling, 15 credits. 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.

    First-frame. Your empty-apron still is 0:00. Prompt paint and camera, not a new fence. Two seconds is the floor. Fifteen is the ceiling. A wrong 15s clip is a longer wrong banking.

    First-and-last-frame. Pin empty apron and empty apron-from-the-next-approved-angle if both stills share geometry. A last frame of a criterium in the street is a morph.

    Video continuation. Extend a take that is already on-model. That is not how you invent the riders.

    If the painted mark is wrong in the still, every mode spends 15 credits animating the typo. Fix the frame, then come back. Crop people out if you do not have a grant. If you need a bike in the apron, shoot the bike in the apron.

    Extend only after the same empty banking holds

    Extension spends on more seconds, not on a different idea. Open the last frame. If you would not freeze it as a still — empty banking, paint true, no invented rider — do not extend it.

    VEO 3.1 Extend Video is the catalog continuation: Google, category extend-video, native audio on, listed max output 720p, duration locked to 7s, 40 credits. Qualities on the record include 720p and 1080p. Aspects: 16:9, 9:16, auto. Image is not required because the input is the clip. You are buying one seven-second continuation per run. Write the next beat that can actually play in seven seconds from the last frame you kept.

    The agent surface is extend a video. The finishing route is extend a video's length. Neither changes the catalog: this Google row still meters 40 credits with audio on. Match the source aspect. An extend that changes 9:16 into 16:9 is a new movie.

    A keeper is a clip whose last frame you would still ship if the video ended there. Drift is not a feature of this row. Do not extend hoping the extra seven seconds will empty a track that already has a generated peloton. Cut or retake first on Wan. Then extend. Native audio is on — if the source was mute on purpose, say so or you will get a bed under a hold you meant to keep dry.

    A generated peloton is a field you did not enter

    A photoreal bunch on the banking is a start list. Those faces are people. An invented one is an extra you did not cast, wearing a race you did not sanction. Youth on the track are a hard stop. Do not “age them up” from a club album to populate a madison.

    Scale is a cone, a distance mark, or a bike already in the photograph. A labelled “generic rider” is still a person-shaped claim if they are on the banking the still did not contain. Empty is the brief. Re-describing “busy velodrome, full field, bunch sprint” is how bodies appear.

    If the still includes a real rider, crop them or reshoot. Background is not a likeness carve-out. Do not generate a rider as proof the track is in use. Session-fill with live racing is a different day, with grants. This leftover is civic architecture of the empty bowl.

    Text-to-video draws riders who are not on the start list

    AI video generator from a paragraph is a new identity. “Cinematic madison, packed velodrome at golden hour, bunch on the banking” rebuilds the bowl and parks a field you did not enter. That is a different venue, and it is a likeness problem wearing a skinsuit.

    Do not send that sentence to a text-to-video row and crop a frame. Make the still. Animate that file. Identity is the I2V choice. Do not “upgrade in the prompt.”

    Hours, open-track times, a hire rate — those are claims you wrote. Wan 2.7 will not typeset them. Overlay after picture lock. Captions transcribe speech; a mute apron has nothing to transcribe.

    The test: would a neighbour say the clip is this track, empty? If not, you generated extras. Plate the painted apron. Extend only after that empty banking still holds.

    FAQ

    Can I text-to-video a packed madison and skip the apron photo?

    You can. You will get a cousin banking and a field you did not enter. The leftover is the apron you photographed. Wan 2.7 I2V requires that image: 2s–15s, 1080p, 15 credits.

    When do I extend instead of generating a longer Wan clip?

    When the last frame is already a keeper — empty banking, paint true. VEO 3.1 Extend Video buys 7s more of that identity at 40 credits, 720p listed, native audio on. It will not empty a track that already has riders. Retake first. Then extend.

    Does Wan 2.7 add a soundtrack to the apron clip?

    Catalog audio is unset. Optional driving audio is a track you supply. It is not native dialogue from a silent prompt. Overlay hours. Score later if you need mix. Do not expect a generated commissaire.

    Can we generate a peloton so the bowl looks in use?

    No. Scale is a mark, a cone, or a bike already in the photograph. A photoreal bunch is a start list you did not enter. Plate the empty banking.