First and Last Frame Video Guide

Give the model compatible endpoints, then describe the visible action that connects them without asking the scene to reinvent itself.

What first and last frames control

In Image to Video, the first image establishes the opening view. On models that accept another frame, the next image specifies the intended ending view. This is different from a collection of appearance references: the order describes time. The prompt should connect those endpoints with plausible action, rather than introduce a new scene that neither image contains.

According to MiniMax’s earlier Hailuo 02 announcement, start and end images can guide a transition between selected endpoints. That is historical background for the technique, not evidence of today’s H3 support. SeaDance support below comes from its current host validation and shared provider mappings. Compatible inputs guide generation; they do not guarantee that every endpoint detail will be reproduced exactly.

Supported models and endpoint prices

All four Seedance entries and both MiniMax H3 entries accept a first frame with an optional last frame. Kling 3.0 accepts exactly one first frame on SeaDance and cannot use a last frame here. Do not place an ending picture in a different input mode and expect equivalent endpoint control. The separate reference-to-video guide explains appearance-reference inputs.

ModelResolutionLast frame5 seconds10 seconds15 seconds
Seedance 2.0 Mini480pOptional4 credits / $0.807 credits / $1.4011 credits / $2.21
Seedance 2.0480pOptional7 credits / $1.4014 credits / $2.8121 credits / $4.21
Seedance 2.0 Fast480pOptional6 credits / $1.2012 credits / $2.4118 credits / $3.61
Seedance 2.5480pOptional11 credits / $2.2121 credits / $4.2131 credits / $6.22
Kling 3.0720pNot supported13 credits / $2.6126 credits / $5.2238 credits / $7.63
MiniMax H3768POptional9 credits / $1.8117 credits / $3.4125 credits / $5.02
MiniMax H3 Max768POptional9 credits / $1.8117 credits / $3.4125 credits / $5.02

Dollar estimates allocate purchased credits at Basic monthly: $29.90 for 149 credits, approximately $0.2007 per credit. They are not separate checkout prices. New accounts start with 0 credits. Check pricing before submitting; each quote below covers one result.

Prices cover one 16:9 video. Adding the optional last frame in Image to Video does not add a reference surcharge. Duration still changes the quote. The multi-image surcharges associated with some Reference jobs do not apply to this first-and-last-frame setup; check the mode as well as the uploaded image count before comparing costs.

Prepare endpoints that describe the same world

Begin with matching composition: keep the same camera height, focal direction, crop and subject scale unless changing them is the purpose of the shot. A vase that begins small and centered but ends cropped at the edge asks for a camera change as well as object motion. If the action is a lid closing, maintain the same viewpoint so the lid can be the main difference.

Match the light source and overall exposure. A warm side-lit opening and a cool overhead-lit ending require an illumination transition that can distract from the intended action. Check shadows, reflections and background lines together. If these are unrelated between frames, prepare a more compatible ending before spending credits. These checks help define a coherent brief; they are not a model-specific technical requirement.

Use clean JPEG, PNG or WebP uploads no larger than 10 MB each. Frame the images for the final 16:9 video and avoid cutting off the object that must move. Keep the product’s shape, color and markings consistent. If you created the ending with an image model, compare it against the opening for accidental changes before treating it as an endpoint.

Write the bridge, not another still-image description

A transition prompt needs the subject’s action, the camera behavior and the intended arrival. “The closed box opens smoothly, ending with its lid raised; the camera remains locked” is more direct than describing a beautiful box twice. If the frames already contain the desired lighting and materials, avoid asking for unrelated restyling. Give the model a clear physical explanation for the difference between them.

Choose 5 seconds for a compact action with nearby endpoints, 10 seconds for a more gradual change, or 15 seconds when a deliberate pause suits the shot. These are creative planning suggestions, not guarantees of motion speed. More time cannot make incompatible geometry physically coherent. If the transition requires a surprising new object, change the endpoint or describe how that object enters the scene.

Four original transition briefs

A package opens

Prepare the same unbranded paper box on the same tabletop, closed at the beginning and open at the end. Prompt: “The paper lid lifts gently backward to reveal the folded tissue inside, finishing in the open position shown in the ending image. The box stays planted, the camera remains fixed, and the warm side light stays consistent.” This suits a simple reveal without adding hands or changing the package design.

A lamp turns on

Prepare matching views of the same desk lamp, with the ending showing the intended light. Prompt: “The lamp gradually illuminates the desk, bringing a warm pool of light onto the notebook while the window light remains steady. Finish at the glow shown in the ending image; keep the lamp’s outline and camera position unchanged.” Check that the endpoint changes brightness rather than replacing the shade or desk.

A curtain reveals a window

Use the same room view with the curtain closed, then drawn aside. Prompt: “The curtain slides gently toward the wall, gradually revealing the garden through the window. Arrive at the open-curtain composition of the ending image, with the furniture stationary and the viewpoint locked.” This suits a spatial reveal with a clear moving surface. Avoid adding a moving camera unless the endpoints already require it.

A flower bud opens

Prepare the same stem and vase with compatible petal colors and framing. Prompt: “The bud slowly opens into the flower shown in the ending image; petals unfold outward while the stem and vase remain still. Keep the camera locked and the soft window light unchanged.” Treat this as a stylized transition rather than a documentary claim about natural elapsed time, and inspect petal continuity after generation.

Troubleshoot the relationship between the images

If the product melts or the room warps, compare the endpoints before adding stricter wording. Reduce differences in perspective and remove conflicting camera instructions. If the ending arrives too abruptly, describe a gradual change and choose a duration that suits it. If identity drifts, prepare a closer matching ending; repeating “exactly identical” cannot repair a different product silhouette in the uploaded image.

Frequently asked questions

Can Kling use a last frame here?

No. Kling 3.0 supports a single first frame on SeaDance. Choose a listed Seedance or MiniMax H3 entry when the ending image must be an input.

Does the last frame cost extra?

Not in this Image to Video setup. The quote depends on the selected model and duration, with no additional charge for the optional ending frame.

Is a second image always required?

No. Supported entries accept a first frame alone or a first frame followed by a last frame. Add the ending only when it expresses a useful endpoint.

Will both images appear perfectly unchanged?

The frames guide the generated endpoints, but exact preservation is not guaranteed. Review important edges, colors, markings and the transition before using the result.

Sources

MiniMax: start and end frame background

Related articles

Appearance references · Image-to-video prompting · Camera movement prompts