Create clips from prompts or reference images. Choose a model, set duration and resolution, and see your credit cost.
Planning prompts
These four briefs are shot-planning recipes for Wan 3.5: product motion, a stage camera path, a first-to-last-frame fashion move, and an outdoor campaign beat. The gallery stills are concept frames for those plans, not proof that a listed model already produced the clips.
Shot workflow
Wan 3.5 keeps one generator for the shot: name the motion, attach frames only when identity or endpoints must hold, read the credit total, then review the clip before you change one control and run again.
01
State who or what is on screen, what moves, what must remain locked, and how the camera should travel. A useful brief sounds like a shot plan: subject, action, lens height, and the beat you will inspect after the run. Name duration intent only in terms the form actually offers, and keep dialogue or native score out of the prompt unless the selected model exposes those controls.
02
Upload a reference still, a first frame, or a last frame when the clip has to start, end, or stay on a known face, product, or set. Skip the upload when you want a net-new text-to-video shot so the workspace is not pulled toward an unrelated still. If you still need a cleaner opening or closing image, prepare it in image mode or on a still-model page, then bring that file back as a frame.
03
Read the credit total on the generator before you start. After the run, watch the clip for identity drift, weak motion, or a crop that misses the brief. Change one thing at a time—the motion sentence, a frame, duration, resolution, or the listed model—then generate again. Stay on the homepage form while you are still comparing settings; do not treat a single pass as a locked delivery.
What Wan 3.5 accepts
Wan 3.5 is the hosted shot desk: a prompt box for motion, optional image references and first or last frames, image and video modes in one generator, a model list that currently includes Wan 2.7 and Wan 2.6, and the credit number shown before you run. Dedicated still URLs exist when you need a frame before the clip, not because this page is a single-model landing page.
The homepage does not hide the selector. Video mode currently lists Wan 2.7 and Wan 2.6; image mode is there when you need a still to reuse as a frame. Switch only among what the dropdown actually shows when look, duration, or input caps matter more than staying on the last choice.
A text-only prompt starts a new shot. The same form takes a reference still, a first frame, or a last frame when the clip must open, close, or stay on a known subject. Image mode is a still-prep path inside the generator, not a second product.
If camera path is the job, name start frame, end frame, and what travels. If on-screen words matter, quote them and then watch those frames after the run. If photography-style materials matter, name surfaces and light so the clip does not drift into a generic move.
The homepage is the planning and review desk for clips. A dedicated image-model URL is for building a still you may later attach as a first frame, last frame, or reference. Do not expect those still pages to output video.
Shot ideas
Use Wan 3.5 when the deliverable is a reviewed clip from a planned shot: product motion, event coverage, continuity between stills, or a campaign beat you can inspect before the next run.
Describe the pack, surface, and camera path so the bottle, label, or kit stays readable while the lens moves. Pin a pack still as a first frame when the SKU must remain recognizable, then review whether the label holds through the move instead of asking the form for a retouched catalog page.
Quote any headline that must appear on a stage screen, lower third, or in-world display. After the run, pause on the frames that carry lettering. If the type collapses, rewrite the quoted words, attach a lettering still as a frame, or switch to a listed model that handles the current duration and crop more cleanly.
Keep wardrobe, face, or interior from the uploaded first or last frame, then change only the camera distance, gesture, or light the prompt names. The workspace is useful here when the job is a controlled move between two known stills, not a new character design.
Plan one outdoor, studio, or in-world shot at a time and compare two listed models on the same motion sentence. Use the clip as a moving storyboard frame: keep the beat if identity and camera path hold, then write the next shot instead of stacking unrelated action into one prompt.
Wan 3.5 vs still-prep pages
Wan 3.5 is the shot-planning workspace. A dedicated image-model page is a still desk for preparing a frame you may later attach. Stay here while the next output you need is a clip; leave only when the next output you need is a still.
Use the homepage when you still need text-to-video, first or last frames, image mode for a quick still, and a selector that currently includes Wan 2.7 and Wan 2.6. Comparison belongs in this form until the shot itself is decided.
Open the matching still guide when you already want GPT-4o lettering, Seedream worlds, or Z-Image portraits as input stills. Those pages prepare frames; they do not replace the generator for video.
The homepage form already accepts uploads for identity and endpoints. You do not need a second video tool to attach a still and run the move.
The generator shows the current model’s credit total for the duration and resolution you selected. Pricing pages explain packs and plans; the button is the source of truth per shot.
Wan 3.5 credits and limits
Wan 3.5 charges credits per generation you start. The generator button is the live price for the selected model, duration, resolution, and mode. Read that number after every change.
This homepage does not advertise a complimentary allotment or an extra hidden tier. Your account balance funds each shot you generate. If the form will not start, check the on-screen total against the balance you actually have, then open /pricing for packs and plans.
Do not quote a blog, a competitor SKU, or a remembered number from an earlier shot. Duration, resolution, and model all move the total. If the number on the button changed after a switch, that new number is the cost of this attempt.
The form only lists durations, aspect ratios, reference slots, and first or last frame controls the current model supports. A missing length or upload slot is a model limit, not a hidden homepage bug. Image mode and video mode also expose different controls.
Commercial terms, refund windows, and credit packs are documented on /pricing. Use this homepage to plan the shot and read the live run cost; use /pricing when you need to add balance, not to guess a per-clip number that the form already shows.
These questions are about the hosted Wan 3.5 video workspace: choosing an input, writing a motion prompt, reading the controls that are actually on the form, reviewing a miss, and checking credit cost before the next shot.
Start here if the shot is still a sentence and you have not decided between text-only, a reference still, or first and last frames.
Use this cluster when you need the live selector, image versus video mode, duration, resolution, or upload slots.
Read this before you change three settings at once or leave the homepage to buy more credits.
Still tools for Wan 3.5 frames
Keep generating clips on Wan 3.5. Open one of these image-model pages only when you need a still to attach as a reference, first frame, or last frame. The cards are concept stills, not video demos, and these tools do not generate clips.
Use the GPT-4o page to prepare a type-forward still, then bring that file back into the workspace as a first frame or on-screen reference. GPT-4o is a still tool here, not a video model.
Pro for high-finish fashion stills; Nano Banana 2 when clothing and weather have to look photographed. Generate the still there, then use it in the workspace as an identity frame.
Open Seedream 4 to build a fantasy or campaign still, then attach that image as a first or last frame for a camera move through the same world.
Z-Image for window-lit portrait stills; Flux 2 when silhouette and contrast lead the frame. Both are still-prep options before a continuity shot.
GPT-4o stills for lettering frames
Concept still: photoreal close-up portrait with cinematic side light, useful as a planned identity frame.
Explore Our Curated Set of Associated AI Models
Nano Banana Pro stills for fashion frames
Concept still: gourmet chocolate dessert with berries and gold leaf, useful when you need a high-finish plate as a frame.
Explore Our Curated Set of Associated AI Models
Nano Banana 2 stills for grounded group frames
Concept still: silver concept car in a concrete atrium, useful as a planned opening frame.
Explore Our Curated Set of Associated AI Models
Seedream 4 stills for illustrated world frames
Concept still: floating island temple at golden hour, useful as a campaign-world first frame.
Explore Our Curated Set of Associated AI Models
Z-Image stills for natural portrait frames
Concept still: moody penthouse living room with a red velvet sofa, useful as an interior frame.
Explore Our Curated Set of Associated AI Models
Flux 2 stills for graphic fashion frames
Concept still: cinematic fighter jet flying through sunlit clouds, useful as a directional frame, not as a clip.
Explore Our Curated Set of Associated AI Models
Stay in the same Wan 3.5 form: rewrite the motion sentence around one beat, keep or drop the frames, switch a listed model only after you name the miss, and read the credit total before you generate again.
Keep subject, locked details, and a single camera path in one sentence. Remove extra scenes this form cannot edit together. If the last clip failed on motion, change the move before you change the subject.
Keep the upload when identity or endpoints must hold. Remove it when you want a net-new text-to-video shot that should not copy the last still. Prepare a replacement frame in image mode if the current still is the thing that failed.
Change among the listed models, including Wan 2.7 and Wan 2.6, when identity, motion, or duration limits of the current family are the problem—not because another site claims a metric this form does not show.
If the number moved after a model, duration, or resolution change, that is the new cost of the shot. Open /pricing only for packs and plans, then return here to run the next beat.
