P-Video-Edit

P-Video-Edit is Pruna’s instruction-based video-to-video editing model. Upload one source video of up to 15 seconds and a text edit instruction; the model returns an edited video. Optional reference images (up to four) guide identity, product, or accessory edits.

Use it to change colorways, materials, environments, on-screen text, lighting, or objects in existing footage without a reshoot. Motion, camera, and timing follow the source clip.

It is optimized for:

  • Instruction-following edits on finished footage

  • Optional reference-guided product and accessory work

  • Cost-efficient draft previews before a full-quality run

    • Price: $0.045 per second of output video (standard) or $0.025 per second (draft)

Not sure how this differs from P-Video-Replace, P-Video-Animate, P-Video, or P-Video-Avatar? See Choosing the right video model below.

Note

When using P-Video-Edit, respect the copyright of videos and images you use as input and of the video you generate.

Pricing: billed per second of the returned video.

Configuration

Price

standard (draft=false)

$0.045 per second of output video

draft (draft=true)

$0.025 per second of output video

Tip

API reference: p-video-edit on Pruna API. Preview: P-Video-Edit playground.

Prompt formula

Fast pass

One principal change. Enough for colorways, simple removals, and first looks.

[video] [prompt]
source clip, up to 15 s, one line: change only X to Y

Locked-in

Name the edit, then name everything that must stay unchanged. Add reference images for product or accessory match.

[video] [images] [prompt]
same source clip every run, optional 1-4 refs for the object or look, change + preserve geometry, motion, camera, lighting
Fast pass: source ↔ output compare; one-line colorway edit on a trench coat.
Locked-in: source ↔ output compare; colorway plus explicit preserve lines for geometry, glass, trim, and camera.

  1. video: required source clip. Maximum length 15 seconds. Motion, acting, timing, and camera movement come from this file.

    • Examples: a campaign hero, UGC take, catalog turntable, or game capture.

  2. prompt: required text instruction. Use one principal change per run. Describe the desired final state, then name what must remain unchanged.

    • Fast pass: “Change only the plain trench coat into a yellow trench coat.”

    • Locked-in: “Change only the SUV body paint to deep metallic red. Preserve the vehicle geometry, glass, trim, headlights, tires, reflections and shadows. Keep the environment, lighting and camera movement unchanged.”

  3. images: optional, up to four reference stills (jpg, jpeg, png, webp). Use when the edit must match a specific product, accessory, or look.

    • Fast pass: omit for color, text, removal, and relighting.

    • Locked-in: pass the SKU or accessory still and say “shown in the reference image”.

Slot

Fast pass (enough to run)

Locked-in (stronger control)

video

Any clean clip ≤ 15 s with the subject in view.

Same clip every run when comparing prompt tweaks.

prompt

One line: change only X to Y.

Change + preserve lines: geometry, motion, camera, lighting, and anything that must not move.

images

Empty.

1-4 refs; name them in the prompt (“the cargo box shown in the reference image”).

Tip

For comprehensive video prompting (motion, framing, atmosphere), see the Video Generation guide.

How does it differ from other models?

P-Video-Edit is a general instruction editor for existing footage. It is not a character-swap model and not a motion-transfer model.

  • P-Video-Replace swaps people in a clip while keeping the shot’s atmosphere.

  • P-Video-Animate moves a still using a driver clip’s motion.

  • P-Video generates new footage from prompts.

  • P-Video-Edit rewrites what is in the shot: color, material, environment, objects, text, lighting.

See P-Video-Animate vs. P-Video-Replace for animate vs. replace. Use this page when the performer and camera should stay and the content should change.

Choosing the right video model

Pruna ships five performance video models. They share the same prediction API, but each solves a different production problem. P-Video-Edit uses the p-video-edit model (Model: p-video-edit header) and requires an existing source video.

P-Video

P-Video-Avatar

P-Video-Animate

P-Video-Replace

Edit (this page)

One-line job

Generate new footage from prompts

Speak from one still (script or audio)

Retarget one still with clip motion

Swap characters in existing footage

Rewrite content in existing footage

You start with

Text prompt (+ optional image refs)

Portrait still + voice_script or audio

Source video + one still

Source video + 1-4 identity stills

Source video + text prompt (+ optional refs)

You keep from the source

N/A (new scene)

Aspect ratio of the still

Motion, timing, camera movement, and optionally audio

Camera, timing, blocking, background

Camera, timing, subject performance; edit the named content

Typical ask

“Make a 10 s product ad in this style.”

“This spokesperson says this line in French.”

“Animate this catalog still using our winning ad take.”

“Put our creator in this UGC b-roll.”

“Turn this silver SUV red and keep the camera move.”

Quick decision guide

  • No source video yet → use P-Video to create the plate, or P-Video-Avatar if you only need a talking head from a still.

  • Footage exists and the hero still should move like the driver → P-Video-Animate. See P-Video-Animate vs. P-Video-Replace.

  • Footage exists and you need different people in the same shot → P-Video-Replace.

  • Footage exists and you need a different color, product, environment, object, or on-screen text → P-Video-Edit.

Tip

Common pipelines: Generate plates with P-Video or shoot live action → P-Video-Edit for SKU, colorway, and cleanup variants → P-Image for reference stills when the edit must match a specific product.

Speed and throughput

Metric

P-Video-Edit

Cost (standard)

$0.045/s of output video

Cost (draft)

$0.025/s of output video

Rate limit

50 requests per minute

Max source length

15 seconds

Draft mode is the faster, lower-quality preview path. Iterate with draft: true, then rerun finals with draft: false.

Key features

Instruction-based video-to-video

One source clip plus one edit prompt. The output follows the source camera, timing, and performance.

Reference-guided edits

Pass up to four stills when the result must match a product, accessory, or look.

Attribute and material edits

Colorways, fabrics, liquids, and packaging variants from one master.

Environment and lighting

Restyle the setting or relight a scene while keeping the subject.

Add, replace, and remove

Attach accessories from a reference, swap a SKU, or clean catalog clutter and on-screen text.

Draft mode

Cheaper preview at $0.025/s before a standard-quality run.

Audio control

save_audio defaults to true so the edited file keeps source audio.

Prompt upsampling

On by default. Turn it off when you need a literal, test-only prompt.

Practical constraints

  • Source video maximum length is 15 seconds.

  • Output length follows the source duration.

  • Reference images: jpg, jpeg, png, webp; up to four.

  • Use one principal change per run. Stacking unrelated edits in one prompt reduces reliability.

  • Draft mode is a preview: faster and cheaper, lower quality.

  • Very fast action, heavy occlusion, or extreme camera motion may reduce consistency.

  • Latency, output resolution, FPS, accepted codecs, and maximum file size are not published as a frozen benchmark on this page.

Examples

Each card shows the source (video), optional reference (images), a source ↔ output wipe compare, and a copy-ready prompt. Full inputs, outputs, and payloads live on prompt_guide/p-video-edit.

Source video
Source ↔ output

Vehicle colorway · attribute editing · video + prompt

Example prompt
Change only the SUV body paint to deep metallic red. Preserve the vehicle geometry, glass, trim, headlights, tires, reflections and shadows. Keep the environment, lighting and camera movement unchanged.
Source video
Reference images Roof cargo box reference
Source ↔ output

Roof cargo box from reference · reference-guided accessory addition · video + image + prompt

Example prompt
Add the roof cargo box shown in the reference image to the SUV. Match its shape, matte-black material and proportions. Keep it rigidly attached and correctly aligned to the vehicle roof throughout the camera movement. Preserve the vehicle body, windows, trim, wheels, pedestrian, environment and lighting.
Source video
Source ↔ output

Campaign environment variant · environment editing · video + prompt

Example prompt
Change the setting into a snowy alpine mountain road at daylight.

API usage

Get your API key from the Pruna User Portal. Use the same header on every call:

-H 'apikey: YOUR_API_KEY'

Step 1: Upload your source video

curl -X POST "https://api.pruna.ai/v1/files" \
  -H "apikey: YOUR_API_KEY" \
  -F "content=@/path/to/source-video.mp4"

Step 2: Create an edited video

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H 'apikey: YOUR_API_KEY' \
  -H 'Model: p-video-edit' \
  -d '{
    "input": {
      "video": "https://api.pruna.ai/v1/files/source-video-abc123",
      "prompt": "Change only the SUV body paint to deep metallic red. Preserve the vehicle geometry, glass, trim, headlights, tires, reflections and shadows. Keep the environment, lighting and camera movement unchanged."
    }
  }'

Synchronous request (draft preview)

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H 'apikey: YOUR_API_KEY' \
  -H 'Model: p-video-edit' \
  -H 'Try-Sync: true' \
  -d '{
    "input": {
      "video": "https://api.pruna.ai/v1/files/source-video-abc123",
      "prompt": "Change only the SUV body paint to deep metallic red. Preserve the vehicle geometry, glass, trim, headlights, tires, reflections and shadows. Keep the environment, lighting and camera movement unchanged.",
      "draft": true
    }
  }'

Optional reference images

curl -X POST "https://api.pruna.ai/v1/files" \
  -H "apikey: YOUR_API_KEY" \
  -F "content=@/path/to/reference-image.png"

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H 'apikey: YOUR_API_KEY' \
  -H 'Model: p-video-edit' \
  -d '{
    "input": {
      "video": "https://api.pruna.ai/v1/files/source-video-abc123",
      "prompt": "Replace the person with the reference",
      "images": [
        "https://api.pruna.ai/v1/files/reference-image-def456"
      ]
    }
  }'

Configuration

Required parameters

Parameter

Type

Description

video

string (URI)

Input video to edit. Maximum length: 15 seconds.

prompt

string

Text prompt describing the edit.

Optional parameters

Parameter

Type

Default

Description

images

array of URI

Optional reference image(s). Up to 4. Supports jpg, jpeg, png, webp.

prompt_upsampling

boolean

true

Enhance the edit prompt.

draft

boolean

false

Faster, lower-quality preview. Billed at $0.025 per second.

save_audio

boolean

true

Save the video with audio.

seed

integer

random

Random seed. Range 0-2147483647. Set for reproducible generation.

Argument recommendations

  • Model header: p-video-edit.

  • video: keep the subject visible; prefer stable exposure and a clip no longer than 15 seconds.

  • prompt: one principal change; add preserve lines for geometry, motion, camera, and lighting.

  • images: use for SKU, accessory, or look match; mention the reference in the prompt.

  • prompt_upsampling: leave true for production; set false only for literal A/B tests.

  • draft: iterate in draft, then rerun the locked prompt with draft: false.

  • save_audio: keep true unless the workflow needs a silent file.

  • seed: set for reproducible A/B tests; change one variable at a time.