Steady footage
Shaky, fast or very busy clips make placement harder. Locked-off or slow shots blend best.
Pikadditions inserts a new subject, such as a person, animal, object or creature, into a video you already have. It keeps the original footage and blends the new element into the scene's lighting and motion. Give it a video, a prompt and, optionally, a reference image of what to add. It's available in Pika's apps and on the Pika API at $0.03 per request.
Pika describes Pikadditions in one line: "Insert any subject into an existing video." It's a video-to-video model. Unlike text-to-video, it builds on real footage, so the new element looks as if it was there when the clip was filmed.
| Spec | Pikadditions |
|---|---|
| Inputs | Source video and a text prompt (both required); reference image of the subject (optional) |
| Extra controls | Negative prompt, seed |
| Output | MP4 at 720p or 1080p (default 1080p) |
| Length | 5 or 10 seconds (default 5) |
| Where | Pika's apps, plus the Pika API and ComfyUI's Pikadditions node |
| API price | $0.03 per request; only successful jobs are charged |
From the Pikadditions model page and API spec. Pika's API docs don't list limits for input video length, format or file size.
Pika's own examples show how specific a prompt can be: where the new element sits, how it moves with the subject, and how it matches the light.
Videos © Pika, from the Pikadditions examples on dev.pika.art.
A steady clip with a clear subject and some open space works best. Your own footage or a clip you have the rights to.
Upload a clear picture of the person, pet, product or object to add, ideally on a plain background.
Say what to add, where it goes, what it does and how it interacts with the scene.
Choose 5 or 10 seconds, and 720p or 1080p.
If placement or scale is off, make the position and size more specific, or add a negative prompt.
Launch-app details as reported by KOCPC. More: Pika iOS app.
Pika's own example prompts are detailed. They describe the subject, its exact placement, how it follows motion and how it matches the lighting. The prompts below are our ready-to-copy starting points; we haven't run each one through Pikadditions.
Name the subject clearly: "a small orange tabby kitten", not "a cat".
Give a position: "on the left of the sofa", "on his shoulder", "behind the counter".
Say what it does and how it reacts: "walks in from the right and sits beside her".
"Matching the warm window light and shadows" helps it blend in.
| Tool | What it does | Use it when |
|---|---|---|
| Pikadditions | Adds a new subject to the scene | Something is missing that you want in the shot |
| Pikaswaps | Replaces part of the video, chosen by a text description or a mask | Something is there that you want changed |
| Pikaffects | Applies a named effect (melt, inflate, squish…) to an image | You want a quick viral effect |
| H3 Flex references | Edits outfits, backgrounds or motion using tagged image, video and audio references, with sound | You want a bigger edit or audio in the same pass |
H3 Flex clips © Pika, from the Pika H3 Flex research post, shown for comparison (not made with Pikadditions). More: Pikaswaps, Pika effects and H3 Flex references.
Submit a job, poll GET /v1/media/jobs/{id} until it's completed, then download the MP4 from output.video.url. Upload local files first with POST /v1/media/uploads, and PUT within 5 minutes.
| Field | Details |
|---|---|
video | Required. Source video URL |
prompt | Required. The edit to apply |
image | Optional. Reference image of the subject |
negative_prompt | Optional. What to avoid |
resolution | 720p or 1080p (default) |
duration_s | 5 (default) or 10 |
seed | Optional, for repeatable results |
Pika's sample job shows a charge of about $0.37, not the listed $0.03, so confirm real charges from billing.charge_micro_usd. More: Pika API guide.
curl -X POST \
https://api.dev.pika.art/v1/media/pika/pikadditions/video-to-video \
-H "X-API-Key: $PIKA_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"video": "https://example.com/my-clip.mp4",
"image": "https://example.com/puppy.png",
"prompt": "Add the puppy from the image sitting beside the woman on the bench, matching the soft light.",
"resolution": "1080p",
"duration_s": 5
}'
Shaky, fast or very busy clips make placement harder. Locked-off or slow shots blend best.
Scale is a common miss. Use phrases like "knee-high", "fits in her hand" or "the size of a car".
One subject, well lit, no clutter. The clearer the image, the closer the match.
Only add real people with their permission, and only edit footage you have the rights to. Label realistic edits as AI when you post them.
See also: Pika watermark and AI labels.