VFX Compose — same model on a Florence rooftop at golden hour

VFX Compose — same model on a sunset cliff

VFX Compose — same model in a fantasy dragon scene

VFX Compose — same model on a neon nighttime rooftop

VFX Compose — same model in a snowy forest at dawn

🎬

VFX Compose

Drop the same subject into any scene.

Same face. Same outfit. Same jewelry. New world. We matte your subject, generate a fresh background from your prompt, blend the lighting, and layer the original pixels back on top. The face you uploaded is the face you get — byte-for-byte.

🚀 Enter Tool 📘 API Docs

What it makes

Six real outputs from the live API. Same source photo, six different prompts, ~21 seconds each. The face, the white blazer, the gold pendant chain, the transparent block-heel sandals — all identical to the source. Only the scene is freshly generated.

Florence rooftop at golden hour with Duomo — Italian Florence rooftop · golden hour · Duomo in distance

Windswept cliff at golden-hour sunset — Windswept cliff · golden-hour sunset · ocean below

Fantasy scene with dragon and full moon — Fantasy scene · floating dragon · full moon · warm clouds

Tokyo neon rooftop at night — Tokyo rooftop · night · neon skyline · cool reflections

Snowy pine forest at dawn — Snowy pine forest · dawn · soft cold light · drifting snow

Cozy living room with warm window light — Cozy living room · warm window light · soft sofa

How it works

A three-stage classical-VFX pipeline. No black-box subject regeneration. No plastic-skin AI redraw.

Stage 1

Matte the subject

BiRefNet (MIT license) cuts an alpha matte at source resolution. Hair wisps, jewelry chains, and transparent shoes are preserved — not flattened.

Stage 2

Generate the scene

SDXL-Inpaint (OpenRAIL++-M license) fills the background around the matte using your prompt. We then re-layer the source subject pixels back on top — so the face stays byte-identical.

Stage 3

Harmonize tone & shadow

Classical Lab-space Reinhard color transfer warms or cools the skin toward the scene's lighting. A procedural contact shadow grounds the subject. No model touches the face.

Why not just "replace the background"?

Background replacement keeps your subject on a flat backdrop. VFX Compose treats the subject like a film element dropped into a freshly built world — with matched lighting tone, a contact shadow, and a scene-aware silhouette dilation so SDXL doesn't paint a halo of the original studio light around you.

Provider	What it does	Cost per image
Bria · Background Generation	SDXL-style scene fill, subject pixels usually preserved	$0.030
Photoroom · AI Background	Scene from prompt, mid-tier matting	~$0.10 effective on Pro plan
Adobe Firefly · Generate Background	Studio-grade fill, subscription gated	~$0.04 in credits
PixelAPI · VFX Compose	BiRefNet matte + SDXL-Inpaint scene + Lab harmonize + procedural shadow. Face byte-identical by construction.	$0.015

Source: each vendor's public pricing page (May 2026). VFX Compose at 45 credits = $0.015 — set deliberately at 2× cheaper than Bria, the cheapest verified direct rival.

Watch it in action

Full walkthrough video is coming. In the meantime, the live tool is ready and the docs page has tested cURL examples.

🎬

Tutorial coming soon

Subscribe to PixelAPI on YouTube for the v2 walkthrough.

📺 Subscribe to PixelAPI on YouTube →

Pricing

Pay-as-you-go credits. No subscription. Free credits on signup. 2× cheaper than Bria's $0.03/image — and our subject pixels are byte-identical to your input.

Per image

45 credits

≈ $0.015 / ₹1.25 per scene · ~21 seconds end-to-end · face byte-identical to source