Prepare a Reference Image for Image-to-Video

Prepare the frame before paying for motion. Choose the final composition, check the details that must remain accurate, and remove defects already visible in the source. A motion prompt is not a dependable way to repair a blurred face, missing product edge or incorrect label.

This guide answers a preparation question: is this image ready to animate? For model selection and generation steps, use the image-to-video guide; for defects that appear after generation, use artifact troubleshooting.

Decide what the finished shot must preserve

Write one sentence describing the shot’s purpose before opening an editor. “Show the full bottle and its cap while the camera approaches” gives you an acceptance condition. “Make this image cinematic” does not tell you whether a cropped cap is acceptable.

Intended result	Inspect in the source	Prepare before upload
Recognizable person	Face clarity, pose, hands and occlusion	Use an authorized, clear photograph with the required features visible
Accurate product	Silhouette, cap, seams and label	Compare with the real product; replace inaccurate imagery
Vertical social clip	Subject position and space for captions	Make a separate vertical composition in an image editor
Environmental shot	Straight lines and details that must stay recognizable	Decide which elements can move and which must remain stable
Only a slow pan or zoom	Whether any new action is actually needed	Consider a still-image pan or zoom in a video editor

A still-image edit is useful when the brief only needs movement across an existing picture. Generating new motion makes sense when the subject or scene needs to change over time. Neither choice removes the need to inspect the final export.

Calculate the crop before selecting a vertical output

A 16:9 image does not automatically become a useful 9:16 composition. For a crop that keeps the full height, calculate:

crop width = original height × target width / target height
retained width share = crop width / original width

For a 1920 × 1080 image cropped to 9:16, the width is 1080 × 9 / 16 = 607.5 pixels, approximately 608. Only 31.6% of the original width remains. A centered crop removes roughly 656 pixels from each side. This is a geometry example, not a measurement of any model’s resizing behavior.

Put a 9:16 crop overlay on the source in your image editor. Check the face, hands, product and required background inside that rectangle. Move the crop if that solves the composition. If two essential subjects cannot fit together, choose a different photograph or make a new composition; simply requesting “keep everything visible” does not settle the conflict.

Expanding the background creates new image content. Inspect that extension for invented objects, repeated textures and broken perspective before using it. For caption placement and delivery framing, see the vertical video guide.

Reject source defects before adding motion

Inspect the original at full size and again at the intended viewing size. Check the eyes, teeth, fingers, product edges, reflected shapes and small lettering. Compare important details with an authoritative photograph or the actual object rather than deciding that a plausible image is accurate.

What you find	Useful next step	What does not establish a repair
Detail exists but is lost in the crop	Reposition the crop or choose another composition	Increasing output resolution
Face or product is already incorrect	Replace or carefully repair the source, then inspect again	Asking for accurate anatomy in the motion prompt
Subject is hidden by another object	Choose a clearer source or a shot that does not need the hidden detail	Assuming generation will recover the true shape
Required text is too small to read	Plan verified text as an overlay in the final editor	Treating generated lettering as approved copy
Source suggests a different motion	Choose a compatible pose or simplify the requested action	Adding several competing motion instructions

Runway’s official image-to-video prompting guide describes the image as the visual starting point and the prompt as the motion instruction, and warns that source artifacts can become more pronounced. That documentation concerns Runway’s workflow; it does not prove identical behavior or controls across VidVid’s models.

Prepare the file for VidVid

As checked on October 5, 2026, VidVid’s upload endpoint accepts JPEG, PNG and WebP image MIME types, with a non-empty file of at most 50 MiB (52,428,800 bytes). This is the upload ceiling, not a recommended file size or a promise that every model accepts every image configuration. Convert an unsupported format in an image editor; renaming a file extension does not convert its contents.

Prepare your crop externally, then open the VidVid studio, select a model with image input, and check its available aspect ratios, durations and resolutions. Model controls vary. A multi-file picker does not establish that every selected model supports multiple reference images. Use the settings actually shown for that model and check any upload or generation error before retrying.

Keep the original separately from prepared versions. Name copies by composition, for example bottle-front-9x16-v1.webp, so you can connect a generated clip to the exact input. Avoid repeatedly exporting compressed copies while preparing alternatives.

Copy this preparation brief

Shot purpose:
Source file and permission to use it:
Prepared file version:
Final aspect ratio:
Details that must remain visible and accurate:
Source defects checked or repaired:
One required subject action:
One camera instruction, if needed:
Selected model / duration / resolution:
Displayed credits and retry ceiling:
Acceptance condition for the exported clip:

Example brief: a full bottle and cap remain visible in a vertical shot; the bottle stays stationary; the camera moves slowly closer; the verified label is added in the final edit. This is a proposed planning example, not a generated result or product accuracy test.

Run one baseline, then identify the failure stage

Start with the prepared frame and a simple action. Record the input version and selected settings. Review the downloaded clip from beginning to end, then inspect the exported edit with its actual crop and captions.

If the missing detail is absent in the prepared image, revise the source. If it disappears during motion, use the troubleshooting workflow. If it disappears only after editing, fix the delivery crop. Keeping these stages separate helps the next attempt answer a specific question instead of spending credits on a vague retry.

For recurring people, continue with the character continuity guide. For budgeting, record credits per accepted clip using the pricing guide.

Evidence and limits

The crop calculation, decision tables and preparation brief are original editorial tools. They have not been validated as a controlled model benchmark. Product upload limits and settings behavior were checked against VidVid’s implementation on October 5, 2026; the selected model and live error messages remain the practical check when using the studio. No success rate, model ranking or AI-search citation is claimed.

Need to create the source still first? Use the image prompt and composition guide to define subject placement, lighting and layout space before preparing it for animation.