Skip to content

Guides

Photo-to-video guide: prompts and troubleshooting

Create a short video from an existing photo. Prepare the image, describe one movement, check credits, and troubleshoot changing faces or sliding backgrounds.

Open studio

Chat and video use credits from your Ofrai balance.

Choose a photo with room to move

Use a sharp, evenly lit photo with a clear subject and a simple background. Leave some space above the head and around the subject. Small faces, blur, and hidden product details give the model less information to work with. Start with your existing photo; generating an extra AI still is not required. Check that the detail you care about most is readable in the source.

Use the photo-to-video page directly

Open Photo to video, upload your photo, and edit the prompt or choose a camera preset. This page submits a video job directly; you do not need an intermediate trip to Image Studio. Check the current model, aspect ratio, duration, and credit estimate before submitting. Review the completed clip and download it. The general Video Studio may offer different models and settings; use the controls actually shown there.

Write one main action

Name the subject’s motion, then the camera and details you want to preserve. For a rain-window portrait, try: ‘The subject looks up slightly as glowing jellyfish drift past the window. Fixed camera, keep the collar, face, and window frame in place, no cuts.’ Swap the jellyfish for something that fits your photo, such as leaf shadows or raindrops. You can edit the text after choosing a preset.

Test a short, simple move first

The scene starts at 5 seconds and 16:9 by default. Judge one movement before combining several. Avoid asking for a turn, a walk, a scene change, and a complex camera move in the same first attempt. Change one instruction between attempts, such as replacing a full head turn with a small glance. This makes results easier to compare. Do not assume controls described for another tool exist on this page.

Budget from the live credit estimate

Check the generation panel before submitting rather than budgeting from a fixed price in a tutorial. Model, duration, and settings affect cost, and every regeneration is a new job. If a job fails, check its status and your account record before retrying. Failed jobs usually refund credits automatically; your account record is the source of truth. A slow task is not a reason to submit it repeatedly.

Troubleshoot drift before trying again

If a face changes, use a clearer photo with less occlusion and reduce head movement. If a product label breaks, reduce the rotation and reflections; use the original photo or compositing when exact lettering is essential. For sliding backgrounds, try a locked camera and explicitly request a static background. These constraints may help but cannot guarantee fidelity. After repeated failures, reconsider the source and action rather than rerunning the same setup.

Review the whole clip before downloading

Check faces, hands, product silhouettes, labels, and backgrounds from beginning to end, including middle frames and the final moment. Confirm that the output format suits its destination. Reference clips show motion ideas, not a promise of tested output from the current model on Ofrai. Confirm rights to source photos before publishing and compare commercial product imagery against the actual item.

To make a photo move with AI, upload it in the photo animation tool and start with one movement.

If your first frame and framing are ready, try the Vidu Q3 Pro first-frame workflow and Seedance 2.5.

For a store listing, use the product photo-to-video generator to turn an approved packshot into a short clip. Check the cost before submitting and compare the finished frames with the original photo.

Generated in Ofrai · GPT Image 2.5 → Seedance 1.0 Pro Fast · 5 seconds

From an idea to a moving story

Try photo to video
Space cat courier

Original fictional character. This example used 21 image credits and 41 video credits. Current prices appear before generation.

Watch the helmet, orange hover scooter, and headlight. All three stay recognizable for five seconds. Road reflections and the shifting background carry much of the motion. For your own photo, choose one thing to move and two details to check.

Break down the jellyfish night scene

Four-second creative demo. The subject looks up slightly while jellyfish and nearby reflections give the shot depth.

The face changes as soon as the clip starts

Start with a sharp photo with little hair or hand occlusion. Keep the camera fixed and try a glance or blink. Compare the first second with the source around the eyes, hairline, and shoulders.

Try the subtle movement preset with your photo

The window frame or background drifts

Ask for slow movement in the lights outside and keep the window frame and subject in place. Check the frame and collar at the middle and end. Change one instruction if you retry, such as replacing a push-in with a fixed camera.

Adjust one instruction and check the estimate

FAQ

Why can an identical prompt give different results?
The source, model, and generation process all affect output. Keep the photo and prompt and change one instruction at a time to compare attempts.
Does asking for unchanged details guarantee fidelity?
No. It is an instruction, not an exact-reproduction guarantee. Faces, product geometry, and lettering still need review.

More guides