Skip to main content

pruna-ai/p-video-2-pro-replace/video-edit

Pruna AI's P-Video 2 Pro Replace recasts up to two people in footage of up to 120 seconds, one reference photo each, while the scene and motion stay in place and the soundtrack is kept by default. Output tiers run from 480p to 4K.

From $0.06/s

Workbench

Input

Video editing
135/2000
Optional directions for selecting the source person, assigning each photo and naming what should remain. For example: “First photo for the rider in blue; second photo for the rider in red.” A solo subject can use an empty prompt. at most 2000 characters
affects price1/1

Drop a video here, or browse · 1 file · Max 90 MB · MP4 · WEBM

  • 1
1 / 1 files
Required source in MP4, MOV or WebM, 1–120 seconds and up to 8K. A longer clip is cut at 120 seconds. Duration and aspect ratio come from this video, with no separate controls. at least 1 characters
Reference images2/2
  • 1
  • 2
2 / 2 files
Required: one or two photos containing one person each. One photo targets the main person; two map from left to right unless the prompt assigns them. Any third photo is ignored. at least 1 items · at most 2 items
affects price
Choose 480p or 768p (default) for direct generation. 1080p, 2K and 4K are upscaled from the 768p result and take longer. These tiers add pixels and sharpness rather than extra model detail; the resolution choice sets the billing rate. API values for 2K and 4K are 2k and 4k.
Keep original audio
On by default to retain the source recording. In our tests, switching it off produced a result without an audio track. This setting does not determine whether audio guides lip movement.
Ignore Source Audio
Off by default, allowing the model to listen to the source. Pruna says this guides the replacement’s lips. Turn it on to stop that guidance; keep_original_audio can still retain the recording.
Seed
Optional integer to support comparisons while changing one input. Reusing it does not guarantee that every frame will match a previous run. no lower than 0 · no higher than 2147483647
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run

Related models and next tasks

Missing the model you need?

PHOTO REFERENCES, EXISTING FOOTAGE

Give an existing performance a new cast, with one or two reference photos.

Key numbers

Source limit
120 S
Footage beyond two minutes is cut at the 120-second mark.
Upscaled output
4K
Pruna upscales a 768p result for the top tier.
People per run
2
Supply one photo for each of up to two replacements.
Kept by default
AUDIO
In our tests, the cycling scene retained its street sound.
Measured MP4 output
24 FPS
In our tests, both measured source durations were retained.
Default resolution
768p
This tier and 480p are generated directly by Pruna.

Before you spend

All modes
  • Permission for each likeness and rights to each photo come first. Do not use celebrities, public figures, acquaintances or photos you have no rights to. Accept the first-run consent notice on the model page and describe edited footage truthfully.
  • The source range is 1–120 seconds. A longer upload is cut at 120 seconds; trim your intended section before submission.
  • Provide up to two photos, with one person in each. A third photo is ignored rather than assigned to another person.
  • 1080p, 2K and 4K are upscales of a 768p result. They take longer and cannot add model-generated detail beyond that result.
  • Treat resolution names as tier labels rather than exact pixel dimensions. In our tests with 16:9 sources, the tiers returned 832×480, 1344×768, 1904×1088, 2536×1450 and 3808×2176.
  • Save generated videos before the 14-day retention period ends. The stored files are then deleted.

What it does

TWO PHOTOS · 768p

Tell each photo which rider to replace

In our tests, the prompt assigned the first photo to the rider in blue and the second to the rider in red. The result showed a man in a white shirt alongside the comic heroine in a brown jacket and cargo trousers. The tram, street and camera movement remained part of the scene.

ONE PHOTO · NO PROMPT

Let the solo performance supply the motion

In our tests, a single reference changed the studio dancer into the woman with dark hair and a black shirt. The request left the prompt empty. Her dance movements followed the source, and the gray studio remained.

SOUND CONTROLS

Separate the soundtrack from audio guidance

The two audio switches serve different purposes. keep_original_audio retains the recording in the delivered file. ignore_source_audio decides whether the model listens to that recording; Pruna says listening lets the replacement’s lips follow it. In our tests, disabling soundtrack retention removed the audio track.

Made with P-Video 2 Pro Replace

  • In our tests at 768p, two assigned photos changed the cycling pair while the street, tram, camera movement and street audio remained.
    5s
  • In our tests at 768p, one photo recast the solo dancer without a prompt; the studio and dance movements stayed in place.
    5s
  • In our tests, a single photo replaced the man in the dark T-shirt with the man in the white shirt at the 480p tier.
    5s
  • In our tests, the 4K tier produced an upscaled version of a 2-second studio dance excerpt.
    2s

Where it fits

  • Casting review on recorded footage

    With the actor’s permission, test a replacement photo against a scene you already filmed and review the look during the existing performance.

  • A new pair for a street scene

    Assign two references to the cyclists by shirt color and examine how each replacement appears through the moving shot.

  • Character styling for a solo routine

    Try a permitted character reference on a dancer’s studio clip and compare hair and clothing throughout the routine.

Model reference

P-Video 2 Pro Replace API reference

P-Video 2 Pro Replace API specifications

Publisher
Pruna AI
Video input
MP4, MOV or WebM; up to 8K
Source length
1–120 seconds; longer uploads cut at 120 seconds
Reference photos
1 or 2; one person per photo
Resolution tiers
480p, 768p (default), 1080p, 2K, 4K; last three upscaled from 768p
Output format and frame rate
MP4, 24 fps in our tests
Output length and shape
Source duration and aspect ratio retained in our tests
Sound
Source soundtrack kept by default; optional removal. Pruna says lips follow source audio unless it is ignored.

How to call the P-Video 2 Pro Replace API

  1. Confirm consent and photo rights

    Obtain the person’s permission for this use and check that you may use the photo. Accept the consent notice shown before your first run on the model page.

  2. Add footage and replacement photos

    Choose an MP4, MOV or WebM lasting 1 to 120 seconds. Add one or two photos, each showing a single person; one photo targets the main person.

  3. Assign the cast and set sound

    For two people, use the prompt to connect each source person to a photo, or use automatic left-to-right matching. Choose a resolution and decide whether to retain sound and let it guide the lips.

  4. Review and save the result

    Compare the replacement’s appearance, performance and soundtrack with your source. Download the finished file within 14 days, before it is deleted.

The short version

Choose footage and photos you have permission to use, make the photo assignments clear, and review the resulting cast in motion. Allow extra time for an upscaled tier and save your MP4 before the 14-day deletion date.

P-Video 2 Pro Replace API questions