Freepik AI Video Guide: Image to Video Models and Settings

Freepik AI is now Magnific. Freepik renamed its AI suite to Magnific; the tools and steps in this guide are the same under the new name.

Freepik does not train its own video model. Its AI video generator (part of freepik.com/ai) wraps several third-party models in a single interface, so you can make a clip with Kling, Hailuo, Runway or whichever others are listed at the time without holding separate accounts. For image-to-video that is handy: one still, several engines, side by side. This guide covers how the generator is laid out, how to prompt it, what the settings do and how to avoid wasting credits.

How the Freepik video generator is organised

Open the video tool and you will see an input area for a start image, a prompt box, a model picker and a handful of settings. Some models also accept an end frame, and some offer text-to-video with no image at all. The model names are the real ones (Kling versions, Hailuo or MiniMax, Runway’s Gen models, and others as they are added), and each has its own credit cost per clip, shown before you generate. The line-up changes often; what matters is that the workflow below is the same whichever model is selected.

Freepik’s video tool shares the credit pool with the image generator, upscaler and the rest of the suite. Video costs many times more per generation than an image, so check the platform’s current pricing page before planning a batch.

Step by step: still to clip

1. Make or choose the still. Generate it in the Freepik image generator or upload your own. Crop to the final aspect ratio first, and leave a little space around the subject for motion.
2. Upload it as the start frame.
3. Write a motion prompt. One action and one camera move: "The model turns her head towards the light and smiles. Slow push-in, static background."
4. Pick the model. A cheaper or faster option for the first attempt.
5. Set duration and aspect ratio. Shortest duration, ratio matching the still.
6. Generate, watch at full size, then step through the first and last second where artefacts appear most.
7. Rerun on a stronger model with the same still and prompt once the motion is right.

Which model for which shot

• Kling. Faithful to the start frame and good with people. The Kling image-to-video guide applies directly, including its mode and duration advice.
• Hailuo (MiniMax). Energetic, cinematic motion; strong camera moves. Good for atmosphere, less predictable for faces.
• Runway. Polished, controlled motion with a filmic feel; a good choice for product and brand work.
• Others as listed. Treat any new entry as a draft engine until it earns its place.

For a cross-tool comparison of these engines, see Kling vs Runway vs Luma for image to video.

Settings at a glance

Prompting for motion

The still already describes the look, so the prompt should not. Name the subject and a single action with a verb. Describe the camera separately: static, slow push-in, gentle pan. Add a pace word; "slowly" and "subtle" prevent the frantic motion that ruins short clips. Say what stays still. Two or three sentences is enough. If the interface offers a prompt enhancer, read what it produced, because that expanded text is what ran, and copy it into your notes if the result is good.

Common problems

• The face changes. Keep the face large in frame, ask for "face stays consistent", avoid fast turns, and prefer Kling for people.
• Warping and melting. Too much motion for the duration. One action, shorter clip, calmer words.
• Text in the still smears. Every video model struggles with lettering. Remove it from the still and add it back in your editor.
• Credits vanish. Expensive model, long duration, several attempts. Draft on the cheap model and only pay for the final.

Keeping a record

Each clip in your Freepik history shows the model and prompt, but the downloaded MP4 carries none of it. When you download a keeper, write the model, prompt, duration and the source still’s filename in a notes file beside it. Clients ask for "that one again, two seconds longer" more often than you would think.

Try it

Start frame: a glass of iced tea with a lemon slice on a sunlit café table, condensation on the glass
Prompt: Condensation drips slowly down the glass. A light breeze moves the napkin. Slow push-in towards the glass, shallow depth of field, background stays still.
Model: the cheapest available for the first run
Duration: shortest option. Aspect ratio: same as the still.

If the drip and the breeze read correctly, rerun the identical inputs on Kling or Runway and keep the one whose motion feels most natural. Change "slow push-in" to "static camera" to see how much of the life was camera versus scene.