h3 flexPika Reference to Video from Images
Bring visual references into a new scene. Add images that communicate your subject, product or visual direction, then describe the action, camera and sound.
H3 Flex is an independent video studio powered by the Pika API.
Sound off by default
Reference-to-Video Example
An official Pika model example for visual inspiration. It is not a guarantee of a matching identity or an exact reproduction of your references.
How to Create a Video with Image References
Choose Your Reference Images
Sign in and add up to nine JPEG, PNG or WebP images, each up to 10 MB. Pick clear, relevant views and use only material you have permission to process.
Describe a New Scene
Explain what should happen, how the camera moves and what the audience hears. Tell the model which visual details matter; images guide a new generation rather than define an ordered timeline.
Review Settings and Credits
Choose 5–15 seconds, a resolution and an aspect ratio. Reference mode uses three times the standard text/image credit rate. Review the displayed total before submitting, then preview and download from My Films.
Reference Images, Limits and Creative Control
References Are Different from Keyframes
Use references to communicate appearance or direction across a new scene. If you need an image to be the starting composition and another to be the ending, choose Image to Video. Adding more references does not guarantee better consistency; compatible, purposeful images are easier to direct.
Supported Inputs and Output
H3 Flex currently accepts image references only. Video and audio reference files are not supported here. You can describe sound in the prompt. Outputs use native 480p or 768p, with larger resolutions upscaled; references do not change that distinction.
An Original Reference Prompt to Adapt
Use the product shape and material in the reference images as visual guidance. Place it on a blue studio set, then make a slow close-up camera move across its surface. Soft side lighting. Sound: a gentle musical phrase and a light metallic click.
Reference-to-Video FAQ
How many images can I add?
Add up to nine image references in the studio. Each upload must be JPEG, PNG or WebP and no larger than 10 MB. Choose complementary views rather than many conflicting scenes; a large image count does not ensure a better result.
Can I upload video or audio references?
Not in H3 Flex at present. This service exposes image references only, even if other upstream API capabilities exist. Describe ambience, dialogue or music in the prompt instead of uploading an audio reference.
Will the model preserve an exact face, character or product?
References guide the generation but do not guarantee exact identity, branding or geometry. Review the full clip, especially small product details and changing angles. Simplify the action and refine your references when a detail is important.
How is Reference to Video different from Image to Video?
Image to Video uses first and optional last frames to anchor the clip. Reference to Video uses images as guidance for a newly described scene. Choose keyframes for a specific starting or ending composition, and references for visual direction.
How are reference generation credits calculated?
The total is duration multiplied by the selected resolution rate and the reference-mode multiplier of three. The studio shows the total before submission. A longer clip or larger output raises the cost; adding a reference is not a separate flat-rate image purchase.
Can I generate audio and download the result?
You can include a sound direction in the scene description. Review the generated soundtrack together with the picture, then download the completed video from your account. Sample videos on this page start muted; that setting does not disable audio generation.