Use video as composition reference

Last updated on Jul 17, 2025

Learn how to use a video as a composition reference when generating video using text descriptions.

If you have a video you’d like to use as a reference for its structure, such as the arrangement of edges, depth, and composition, you can upload it when generating a new video using the Text to video feature. By combining this reference video with your text prompt, the final generated video maintains the composition of the original while aligning with your creative intent and vision.

Before you begin:

Have a reference video ready that is 5 to 10 seconds long and under 200 MB in size. For resolution, you can choose from 540p, 720p, or 1080p, with 1080p recommended for the best results.

  1. On the Firefly homepage, select Text to video.

  2. In the General settings section, select Model as Firefly Video.

  3. Select the resolution from the following available options:

    • 540p
    • 720p
    • 1080p
  4. Select the aspect ratio for the video generation from the available options:

    • Widescreen (16:9)
    • Vertical (9:16)
    • Square (1:1)
  5. Frames per second will be the default 24 FPS, and Duration will be 5 seconds.

  6. In the Composition section, upload a reference video whose edges and depth you want to transfer to the generated video.

    In the Composition section, under Reference, a video is uploaded. There are option to Reset to remove the reference video.
    Use a video as a reference for composition to guide the layout and visual style when generating a new video from text descriptions.

    Consider the following while uploading a composition reference video:

    • Videos must be 5 to 10 seconds long. Longer videos will be trimmed to the first 5 seconds.
    • File size should be 200 MB or less.
    • Use 1080p resolution for best results.
    Note

    The Composition section will be unavailable if you’ve added an image as a keyframe under the Frame option for video generation.

  7. Add a text description for the video you want to generate in the Prompt field.

    Tip
    • The best prompts when using composition references are descriptive. For example, 'The yellow ball bounces across the screen,' rather than command-based prompts like 'Change the red ball to a yellow ball.'
    • You can also use the Enhance prompt option next to the Generate button to turn on prompt enhancement whenever you finish writing a prompt. Once you enable this feature, Firefly will improve your original prompt for you.
  8. If you want to specify the visual style for the generated video, you can either write a descriptive prompt that clearly outlines everything in the video, including the desired style or look, or choose a style preset from the Style section.

  9. If you want to, use the Seed option in the Advanced settings section to add a seed number that helps start the process and controls the randomness of the video model's creation. Using the same seed, prompt, and control settings, you can generate similar video clips.

  10. Select Generate to trigger the video generation.

    Note

    Learn how generative credits are used each time you generate a video or use the sparkles   icon next to the Generate button to view the credits usage details.

  11. Once the video is generated, you have the option to preview and download it.

  12. If you want to add sound effects to the video that you have generated using text prompts or use your own voice recording, hover over the generated video and select Generate sound effects.

    The generated video is in preview mode, and when you hover over it, the Generate sound effects option becomes active. This option is highlighted to indicate that the video can be used on the Generate Sound Effects page for adding sound effects.
    Generate sound effects for the generated video using your voice as a guide or by using simple text descriptions.

The generated video is also saved to the Generation history page. To view all your generated videos, use the Files option in the top menu bar.