Skip to content

Early Black Friday: +100 bonus credits on the First Film Pack.Early Black Friday A one-time Black Friday bonus, one per account, at the standard US$ price.

Get +100 bonus credits
FXI Video Models · Reference-to-Video

P-Video 2 Pro Reference

Lock a subject’s look across a generated clip from reference images, video and audio, not just a single first frame.

Upload up to 9 reference images, 3 reference video clips and 3 reference audio clips in any combination, then write the shot: the camera move, the setting, the performance. The references carry the look; the prompt directs the action.

Credits
from 19 / clip
Duration
5–15s
Resolution
480p–1080p
P-Video 2 Pro Reference · 16:9 · Made with FXI
Input → Output

One reference in, a directed clip out.

The subject on the left is the only reference this clip used. The prompt did the rest: the camera move, the setting and the performance.

Reference image
Reference image used as input for the generated clip below
Generated clip
Made with FXI
How To

How to use P-Video 2 Pro Reference in FXI.

  1. 1
    Open the composer

    Go to Generate and select P-Video 2 Pro Reference from the video model picker, or follow the gold button on this page, which preselects it for you.

  2. 2
    Upload your references

    Attach up to 9 reference images, up to 3 reference video clips (15s combined) and up to 3 reference audio clips (15s combined). Any combination is fine; none of them are required on their own.

  3. 3
    Write the action, not the look

    The references already carry the subject’s look. Use the prompt to direct the shot itself: the camera move, the setting, the performance.

  4. 4
    Pick resolution, duration and mode

    480p, 768p or 1080p; 5 to 15 seconds; Speed for a fast balanced render, or Cost for the cheapest render when you are still iterating.

  5. 5
    Render

    The clip comes back with generated sound by default, or carries one of your reference audio clips through as the soundtrack if you turn that on.

Prompt Tips

Write the action. The references carry the look.

Describe the shot, not the subject

Skip physical description entirely; the references already lock the subject’s look. Spend the prompt on action, camera movement and setting instead.

Use clean, well-lit references

A sharp, evenly lit reference image gives the model the clearest signal. A blurry or dark reference carries through to the output.

Match the aspect to your references

Adaptive aspect ratio frames the clip to match your reference images instead of forcing a fixed 16:9 or 9:16 crop.

Cost mode for drafts, Speed for finals

Iterate on a shot in Cost mode, then switch to Speed (the default) for the render you keep.

Inputs & Limits

What P-Video 2 Pro Reference accepts.

Reference images
Up to 9. JPEG, PNG or WebP.
Reference video
Up to 3 clips, 2–15s each, 15s combined. MP4 or MOV.
Reference audio
Up to 3 clips, 15s combined. MP3, WAV, FLAC or M4A.
Duration
5 to 15 seconds, in 1-second steps.
Resolution
480p, 768p (default) or 1080p.
Mode
Speed (default, balanced) or Cost (cheapest, slower).

Credits scale with resolution, mode and duration, plus how many references you attach. The composer shows the exact price before you render, never less than what you are charged.

Frequently Asked

P-Video 2 Pro Reference, answered.

What does reference-to-video do?

It locks a subject’s look, and optionally their voice and motion, across a generated clip, using reference images, video or audio as the source instead of a single first frame. The prompt then directs the action rather than describing the subject.

What file types can I upload as references?

Images: JPEG, PNG or WebP. Video: MP4 or MOV. Audio: MP3, WAV, FLAC or M4A. You can mix all three types in a single request.

How many references can I use?

Up to 9 reference images, up to 3 reference video clips (15 seconds combined) and up to 3 reference audio clips (15 seconds combined), all in the same request if you want.

What duration and resolution does it render at?

Clips run 5 to 15 seconds at 480p, 768p or 1080p. Speed mode renders a balanced default; Cost mode renders slower for a cheaper credit price, useful while you are still iterating on a shot.

How much does a clip cost in credits?

Credits scale with the resolution, mode and duration you choose, plus how many references you attach, starting from 19 credits for the cheapest tier with no references. The exact price is shown in the composer before you render.

Looking for a simpler first or last frame transition instead? Try P-Video 2 Pro, or read the P-Video prompting guide for frame-chained shots.

Direct your next shot with P-Video 2 Pro Reference.

Upload your references once, then direct as many shots as you want from the same look.