What aspect ratio does SpicyFrame generate?
Every video is vertical 9:16, the shape used by TikTok, Instagram Reels and YouTube Shorts. Square and landscape output are not available.
SPICYFRAME · PRACTICAL GUIDES
Create vertical AI video from photos: crop a 9:16 reference, frame motion for a phone screen, keep captions clear of buttons and export a clean MP4.
5 seconds · 480p · $3.99. Pay per video, no subscription required.
Updated

A public SpicyFrame example illustrating the guide. Results depend on the photo and prompt.
To make a vertical AI video from photos, remember that SpicyFrame always generates 9:16 video, so the job is to give it a photo that suits a tall frame. Crop or choose a portrait-oriented reference, keep the subject near the centre with room above the head, and request one action that reads clearly on a phone. Then pick 5, 10 or 30 seconds and 480p, 720p or 1080p and download the MP4.
The reference sets the starting composition. A portrait-oriented photo already contains what a 9:16 frame needs: the subject from head to the required body point, plus some floor and sky. A wide landscape photo forces the model to reframe it, either cutting the sides away or inventing what lies above and below the original picture. Invented areas are where backgrounds wobble and proportions drift. You get steadier, more predictable vertical AI video when the photo and the output share the same shape.
Open the photo in your phone’s editor or any image app and choose the 9:16 crop preset. For a portrait, leave a little space above the head and keep the chin above the lower third. For a full-body shot, keep the feet in frame with a small margin and leave space in the direction the subject will walk. Crop from the original, not a screenshot, and export as JPG, PNG or WEBP under 10 MB; 1080 × 1920 is a good target when the source is large enough. Upscaling a small crop adds no real detail.
You have three options. First, crop tightly around the subject to 9:16; this works when the subject is large and sharp in the original. Second, upload the wide photo as it is and let the model compose the vertical frame; this is less predictable, so keep the subject central and test five seconds first. Third, split the roles: use a vertical crop of the person as @Image 1 and the full wide photo as @Image 2 for the setting, then write “@Image 1 as the subject, @Image 2 for the location and light.”
Vertical frames favour movement along the depth and height of the image rather than across it. A subject walking toward the camera, a slow push-in on a face, a crane up from shoes to face or a tilt from the sea to the sky all use the tall frame well. A long sideways pan leaves the subject only briefly in a narrow window. Keep the face in the upper-middle area through the whole clip, and prefer one person with a clear silhouette over several people competing for a small width.
On TikTok, Reels and Shorts, usernames, captions and buttons sit over the bottom of the frame and along the right edge, and some interfaces add elements at the top. Keep the face and key action in the central area, and avoid placing important detail in the lowest part of the frame. Add titles and captions in your editor or the app rather than in the prompt: text generated inside the video can change between frames, while text added afterwards stays sharp and correctly spelled.
Walk toward camera: “@Image 1, vertical full-body framing, takes three slow steps toward the camera and stops, gentle backward tracking shot, soft daylight, keep the outfit unchanged, footsteps.” · Reveal: “@Image 1 standing on a terrace, slow crane up from the shoes to the face, warm sunset light, light breeze, preserve the dress.” · Portrait: “@Image 1, head-and-shoulders vertical portrait, turns toward the camera and smiles, very slow push-in, window light.” · Scenery: “@Image 1, slow tilt up from the waves to a pink evening sky, static position, gentle surf.” · Nightlife: “@Image 1 leans against a neon-lit wall, glances at the camera, slow push-in, haze, distant city hum.”
Five seconds gives one visual beat for a loop or transition; ten seconds lets a short action settle; thirty seconds holds a slow continuous scene you can also trim into shorter cuts. Medium 480p costs 399, 799 or 1,999 piments and High 720p costs 799, 1,599 or 4,499 piments for 5, 10 or 30 seconds (100 piments = €1). Ultra High 1080p costs 1,999, 3,499 or 9,999 piments. Test framing and motion in Medium, then render the version you publish in High. Ultra High 1080p is also available. SpicyFrame does not offer square, landscape or native 4K output.
Watch the finished MP4 full-screen on a phone, because that is how viewers will see it. Check that the head is never cut off by the top edge, that the action stays in the central area and that the end frame is usable. Then open the clip in the app you plan to post to and look at it with the interface visible; some feeds and profile grids show a cropped preview, so a centred subject is safer. Listen to the ambient audio before adding music.
Use only photos you own or have permission to transform, featuring adults; an identifiable real person must be 18+ and have agreed to the AI transformation. SpicyFrame does not publish to Instagram, TikTok or YouTube for you, and each platform applies its own content and upload rules, including labels for realistic AI-generated video. Glamour and fashion moods are welcome in the studio, but explicit content is not produced, and platforms may be stricter than the studio about suggestive content.
| Shot | Subject placement in 9:16 | Camera move that suits it |
|---|---|---|
| Close portrait | Eyes in the upper third, space above the head | Very slow push-in |
| Waist-up | Centred, head below the top safe area | Static or gentle handheld |
| Full body | Head to feet with a small margin | Backward tracking as the subject walks |
| Scenery | Horizon high or low, not mid-frame | Slow tilt up or crane |
| Product | Centred with clear space around it | Locked camera, moving light |
Every video is vertical 9:16, the shape used by TikTok, Instagram Reels and YouTube Shorts. Square and landscape output are not available.
Yes, but the model must reframe it for a tall frame. Cropping the subject to 9:16 first, or using the wide photo only as a setting reference, gives steadier results.
9:16 is the shape; 1080 × 1920 is a common recommended size for that shape. SpicyFrame generates 9:16 video in 480p, 720p or 1080p.
Medium is native 480p, High is native 720p and Ultra High is native 1080p, all vertical. Use 720p for clips you publish full-screen.
In the central area, away from the bottom and right edges where app buttons and captions appear. Add text in your editor rather than in the prompt.
Five seconds suits a loop or transition, ten seconds a short action and thirty seconds a slow continuous scene. Choose by the action, not by default.