9 live image models. Video models counted separately.

PHOTO TO VIDEO · PRACTICAL GUIDE

How do you prepare a photo for AI video?

Start with a clear subject, room for movement and the right frame. Check the photo before you spend generation credits.

Crop, resize and compress locally. No sign-in or credits for these tools.

Existing Firestarter exampleSeedance 2.5 · 8s · 480p · AI-fictional adult

This retained clip shows an upper-body composition, not a full-body framing test. See the Firestarter page for its current generation settings.

START WITH THE FILE YOU ALREADY HAVE

Three checks before you generate.

Before opening a tool: use a still JPG, PNG or WebP up to 5 MB and 20 megapixels. If your file is larger, first export a smaller copy on your device. HEIC and animated files are not supported; changing the filename extension does not convert them.

1. Match the photo to the action.

For a small head turn or expression, a clear head-and-shoulders photo can be enough to describe the shot. For a full-body action, choose a wider photo where the head, hands and feet are visible, with space around them. Avoid heavy blur, a covered face or several competing subjects when you want one person to move.

These are preparation checks, not a promise that a model will preserve identity or anatomy. A prompt cannot make cropped-out details reliable. If the original frame is too tight, choose another source photo instead of stretching it or expecting missing limbs to be reconstructed accurately.

2. Choose the video shape first.

PromptToVisual’s image-to-video studio offers square 1:1 and landscape 16:9. The Firestarter template uses 16:9. Match the photo to the selected output, rather than assuming every social-video format is available.

In Crop, choose 1:1 or 16:9, keep zoom near 1× to retain more of the source, then adjust horizontal and vertical position. Inspect the whole frame before downloading. Cropping removes edges; it does not add background. If the chosen shape cuts off the action, use a wider original. The crop tool also has Original and 4:5 options, but no dedicated 9:16 preset.

3. Export, reopen and inspect.

Use Resize only if you need different dimensions, with Keep proportions enabled. Use Compress when an accepted input still needs a smaller export. Lower quality in small steps and inspect the face, hands and fine edges. Compression can soften detail; an already optimized image may even export larger.

Download the edited image, then reopen it on your device. Confirm that it is upright, sharp enough for the intended framing, and still within the video uploader’s 5 MB limit. Keep the original separately. Files do not transfer automatically between these tools and the video studio.

The linked local tools process photos in your browser. Submitting an AI generation is a separate action that uploads the image for processing. Use a photo you have permission to submit.

Describe one clear movement.

Start with a movement that fits the visible composition. For a portrait, an example instruction is:

Keep the camera fixed. The person slowly turns their head toward the camera and gives a small smile. Keep the background still and the framing unchanged.

This is a writing example, not a tested prompt or the prompt used for the clip above. Review the result for face changes, extra fingers, distorted edges and unexpected camera movement. Photo preparation does not guarantee an exact dance, identity match or motion sequence.

Match the photo to the format, not just the theme.

A close portrait can support a makeup reveal, while an overhead outfit-change shot needs a wider source with more room around the body. A character skit needs space for its main prop and camera move. These are different framing jobs even when all three are described as Halloween videos.

Compare all three formats before choosing a tool →

The linked format guides distinguish the external 9:16 templates from PromptToVisual’s current four-second square or 16:9 form.

Frame a one-photo transformation around the detail that changes.

A whole-body pet transformation needs the animal’s outline, paws and room around the frame. A close flower action needs a clear nose, eyes and one visible hand or the pet’s front paws. These are different source-photo jobs even though both begin with one image.

Compare both one-photo transformation formats →

The linked guides separate the external 9:16 templates from PromptToVisual’s current four-second square or 16:9 form.

Choose the source for the output job, not the trend name.

A chest-up selfie can support one accessories edit. A rude-cat action needs the whole pet and its paws. A seven-look doorway montage needs a sharp identity reference plus either a fixed external template or several original clips recorded from the same camera position.

Compare the still edit, short action and montage formats →

The comparison distinguishes external 9:16 presets from PromptToVisual’s separate image editor and four-second square or 16:9 video form.

Prepare the source for the movement or close-up that defines the format.

The Lil T Jones and Fuzzy Dream dance formats both need a full-body source, but the low Fuzzy Dream squat needs the widest floor and side space. The couple-eye still is a different job: it needs two sharp close face photos with both adults’ permission and enough visible eye detail to inspect the result.

Compare source count, framing and available workflows →

The exact external presets and PromptToVisual’s general image-combining and four-second video tools have different inputs and outputs; choose by the job you actually need.

Match the photo to its role in the shot.

A duo swap needs two ordered subjects with enough body visible to hop. A hallway reaction needs one sharp upper-body portrait. A hand-cover makeup reveal needs a clear close face and room for the hand to cross the camera. These are different source jobs even though each begins with a photo.

Compare the photo-to-trend source roles → · Compare Halloween reveal formats →

The linked external presets use vertical multi-shot sequences. PromptToVisual’s current form uses one first-frame image for a separate four-second square or landscape clip.

Separate two-role scenes from a one-photo original shot.

Passenger Princess and Coffee Making use two ordered subjects on the linked external services. PromptToVisual’s current video form accepts one first-frame image, so its alternative must be a different single-subject four-second shot. An original Red Riding Hood scene also starts from one portrait and should not copy a creator reel or soundtrack.

Compare the two-role and story formats →

Use original or licensed audio, and keep photos, characters and reference footage within the permissions you actually hold.

Review once, then submit once.

  • Open the studio and choose the downloaded photo again.
  • Check the selected model, output settings and displayed credit cost. Video requires a verified account and sufficient paid credits; free trial image credits do not apply.
  • Submit once. If the response is uncertain, check History or the existing task’s status before starting another request.
  • After a successful job, play the result and download the private MP4. Check its actual content before deciding whether another generation is needed.

If the file is rejected, fix its format or size before submitting. If a job fails, read its recorded status; a finished video that misses your creative direction is different from a technical failure. Another generation is another request and can consume credits.

Updated November 8, 2026. Tool limits follow the linked PromptToVisual tools; available video settings are shown in the studio.

All practical guides
Checking image studio availability…