Reference to Video

Add reference images of a person, pet, product or style, describe a new scene, and SceneWish generates a video that aims to keep them recognizable.

0 / 7

Reference images

Add at least one reference and describe the scene.

Reference to video examples

See how reference subjects appear in a new scene. Official external examples and SceneWish recordings keep their source labels.

Video examples

A pet becomes a cockpit character
A beagle takes the pilot seat
A border collie in the cockpit
Two references: a person and a cat
A person presents a product
A familiar face at play
A city-street selfie
A dog meets a cat in the park
Two references meet on a street
A portrait and outfit in a reading scene
A ride through the fields
Boxing with a panda
An unexpected penguin guest
Hanfu and an umbrella on a stone bridge
Breakfast in a new room
A familiar face on the subway
A new day at the office
Walking a dog under cherry blossoms

The same reference subjects, different results

A panel comparison from the HunyuanCustom research page uses the same person, cat and drawing prompt. Model versions and complete settings are not published.

Shared input and prompt

A person uses a paintbrush to work on a cat picture at a drawing board.

HunyuanCustom — cat in the picture
Kling, version unspecified — cat beside the board
Vidu, version unspecified — cat near the easel

External research comparison, cropped into model panels with the original labels retained. “Ours” means HunyuanCustom, not SceneWish. The suggested setup uses Kling v3 Omni and is not an exact reproduction. View source

Use this setup

Which model for reference to video?

Reference limits differ a lot between models, and some also accept reference videos or audio. These are the ones to look at first; the picker lists the rest.

  • Seedance 2.0 Fast

    ByteDance

    You provide
    Images: up to 9Optional videos: up to 3, 2–15s eachUp to 15s in totalOptional audio: up to 3At least one image, video or audio reference
    Length
    5–15s
    Resolution
    480p / 720p
    Choose it when
    You want up to 9 references plus optional video or audio at a lower cost.
  • Seedance 2.5

    ByteDance

    You provide
    Images: up to 30Optional videos: up to 10, 2–30s eachUp to 30s in totalOptional audio: up to 10At least one image, video or audio reference
    Length
    4–30s
    Resolution
    480p / 720p / 1080p
    Choose it when
    Your scene needs many references, up to 30 images.
  • Veo 3.1 Fast

    Google

    You provide
    Images: 1–3
    Length
    8s
    Resolution
    720p / 1080p / 4k
    Choose it when
    You have one to three images and want a fixed 8-second clip.
  • A border collie in the cockpit

    Kling v3 Omni

    Kuaishou

    You provide
    Images: 1–7
    Length
    3–15s
    Resolution
    720p / 1080p / 4k
    Choose it when
    You use a few images and want clips up to 4K.
  • Wan 2.7 Reference

    Alibaba

    You provide
    Images: up to 5Optional videos: up to 5, 1–30s eachUp to 150s in totalOptional audio: up to 5At least one image, video or audio reference
    Length
    2–15s
    Resolution
    720p / 1080p
    Choose it when
    You want to mix up to 5 images with reference videos.
  • Vidu Q4 Preview

    Vidu

    You provide
    Images: 1–15Optional audio: up to 3
    Length
    3–16s
    Resolution
    540p / 720p / 1080p / 2k / 4k
    Choose it when
    You have a large set of images, up to 15.
  • Gemini Omni Flash Preview

    Google

    You provide
    Images: 1–16
    Length
    Chosen by the model
    Resolution
    720p
    Choose it when
    You want to use up to 16 images and let the model decide the length.

How to make a reference to video clip

Gather clear references for reference to video, then describe a scene that uses them.

Three steps

1. Add clear references

Upload or choose images of the subject. One subject per image, a visible face or label and different angles help the model recognize it.

A pet becomes a cockpit character — 1. Add clear references
Media from the example above. Use your own inputs; see the example for its source.

Credits for reference to video

Cost follows the model, length and resolution, and reference videos can add to the quote. The number shows beside Generate before you submit, so you can trim the reference set first.

Starter

$7.99/mo

4,800 credits

For occasional short videos

  • 4,800 credits
  • Text and image video creation
  • Private video playback and downloads
  • Credits returned for failed generations
  • All annual credits arrive after payment and expire at the end of the annual billing period.

Pro

$15.99/mo

9,600 credits

For regular video creation

  • 9,600 credits
  • Text and image video creation
  • Private video playback and downloads
  • Credits returned for failed generations
  • All annual credits arrive after payment and expire at the end of the annual billing period.

Max

$32.49/mo

20,000 credits

For higher video volume

  • 20,000 credits
  • Text and image video creation
  • Private video playback and downloads
  • Credits returned for failed generations
  • All annual credits arrive after payment and expire at the end of the annual billing period.

Reference to video: common questions

How references work, how many to add and what to expect.

You add reference images, and sometimes reference videos or audio, so the model knows what your subject looks like or sounds like. Your prompt then describes a new scene. The result is a new video, not an animation of one of your photos.

Your subject, a new scene

Where should your reference to video subject go next?

Add a few clear references, describe a scene and see the video before you download it.

Make a reference video

Selected examples from this tool page. External clips retain their original source; results vary with model and input.