Image to Video Options

Learn how to control Image to Video generation.

Prompt

The Image to Video API accepts text prompts to define the result. Prompts can be generated automatically using Autoprompt AI or provided manually.

Autoprompt AI: Automatic prompt generation

Autoprompt AI analyzes the input product image and generates a text prompt tailored to create a contextually relevant video. This is useful for automating image to video generation, especially with large batches of images.

How to use it: Set the generate parameter to true within the prompt object in the options section of your API request.

{
    "options": {
        "prompt": {
            "generate": true
        }
    }
}

You can influence Autoprompt AI by providing guidelines:

{
    "options": {
        "prompt": {
            "generate": true,
            "guidelines": "zoom out, glitter particles"
        }
    }
}

Custom prompts: Manual control

For even more precise control over the animation and visual elements, provide your own prompt. Custom prompts must be between 3 and 5000 characters.

Example:

{
    "options": {
        "prompt": "The camera slowly rotates left to right around the burger with natural and realistic motion. Steam rises gently from the top, drifting upward. Bacon edges shift slightly from the heat. Melted cheese softens and begins to drip slowly. A thick drop of sauce slides down the side. The grilled patty glistens as it turns, showing juicy texture. Lettuce and tomato stay fresh and still. The bun gives off a soft heat shimmer. As the camera moves, the full burger looks hot, fresh, and ready to eat. The background shifts gently with it, keeping the burger centered. Soft, out-of-focus lighting moves across the backdrop, creating a rich, appetizing atmosphere."
    }
}

Sound

Generated videos include an audio track. The video model creates sound together with the visuals: ambience, sound effects, and speech when the prompt calls for them. Audio generation cannot be disabled at the model level, and every output file contains an audio stream.

Default behavior with Autoprompt AI

When you use automatic prompt generation ("prompt": {"generate": true}), Autoprompt keeps audio to a near-silent minimum unless your guidelines ask for sound. To get a rich sound layer, describe it in the guidelines, for example: "slow push-in with a soft fabric rustle" or "she says: this serum changed my routine".

Minimizing sound in custom prompts

If you provide your own prompt and it says nothing about audio, the model improvises its own sound, which may include music or speech. To keep a clip quiet, append this to your prompt:

AUDIO: Near-total silence. Barely audible neutral room tone at extremely low volume. No dialogue, voices, music, sound effects, footsteps, wind, or environmental sounds.

This reliably minimizes audible content, but it is best effort: it does not guarantee a silent result, and the file still contains an audio track.

Guaranteed silence

For guaranteed-silent output, remove the audio stream after downloading the video:

ffmpeg -i video.mp4 -c:v copy -an video_silent.mp4

This copies the video stream without re-encoding, so it's lossless and takes a fraction of a second per clip.

Generating silent videos at scale? We can handle audio stripping inside a managed pipeline as part of an enterprise plan. [Contact us →]

Duration

The duration parameter controls how long the generated video will be. It can be set to 5 seconds or 10 seconds.

{
    "options": {
        "prompt": {"generate": true},
        "duration": 5
    }
}

Possible values

Parameter	Type	Range	Default value
prompt	string or object	3 - 5000	is required
duration	integer	5, 10	5