Skip to content

Public API · accounts open

Fast, affordable image AI, ready to use.

PixelDepot builds, benchmarks and hosts production image AI models, and serves them through fast, secure and consistent APIs, metered per image so you pay only for what you process. Every model is rights-reviewed for commercial use before it ships.

We build it. We test it. We host it. You call one API.

  • Fast and benchmarked
  • Priced per image
  • Rights-reviewed for commercial use
  • Securely hosted
curl https://pixeldepot.ai/v1/face/analyze \ -H "Authorization: Bearer $PIXELDEPOT_KEY" \ -F "image=@guest.jpg"

Detected

face_1 · 0.9997
synthetic sample

1536 × 2048
184 ms · 1 face

{ "faces": [{ "id": "face_1", "confidence": 0.902, "bounding_box": { "x": 390, "y": 706, "width": 781, "height": 780 }, "attributes": { "estimated_age": { "range": "18-64", "estimate": 26 }, "perceived_gender_presentation": { "label": "Female", "confidence": 0.994 }, "has_glasses": { "label": "yes", "confidence": 0.968 } } }], "meta": { "release_status": "experiment" } }

The problem

Image AI is moving faster than anyone can put it in production.

A convincing demo reel does not tell you:

  • Whether the model is truly production-ready.
  • Whether the rights behind it allow commercial use.
  • Whether it performs reliably outside its demo images.
  • How to host, optimize, scale and maintain it.

PixelDepot handles all of this.

How PixelDepot works

From research to production API.

  1. Research

    We follow the state of the art in image AI and select the capabilities worth shipping.

  2. Verify

    We review the rights behind every model: code, weights, training data and dependencies.

  3. Evaluate

    We benchmark accuracy, speed, hardware requirements and real-world performance.

  4. Improve

    When necessary, we train, fine-tune, optimize or combine models.

  5. Deploy

    We securely host the selected models and expose them through one consistent API.

The catalog

One API for fast, affordable image AI.

Start with Face Analyze, then expand into face detection, background removal, restoration, upscaling, image transformation and other carefully selected services.

  • /v1/face/analyzeLive

    Face Analyze

    A bounding box per face, then estimated age, perceived gender presentation, apparent ancestry and whether the face is wearing glasses, each with a calibrated confidence and a flag saying whether it cleared its validated bar.

    Apparent-attribute features are unavailable for EU processing.

    PixelDepot Face v1

  • /v1/face/detectPrivate beta

    Face Detect

    Geometry only, no attributes, no retention. The cheapest way to count and crop faces in a crowd shot.

    PixelDepot Face v1

  • /v1/image/matteLicense review

    Background Removal

    Alpha matte and cutout for product and portrait images, tuned for hair and event lighting.

    Engine named at approval

  • /v1/image/enhanceLicense review

    Restore & Upscale

    Denoise, sharpen and 2x to 4x upscale for print output from booth-quality captures.

    Engine named at approval

  • /v1/image/transformLicense review

    Image to Image

    Style and scene transforms with a fixed prompt contract, so results stay predictable across a run.

    Engine named at approval

  • /v1/face/matchPlanned

    Face Recognition

    Galleries and 1:N matching, gated behind biometric-privacy and commercial-licensing sign-off.

    Engine named at approval

Nothing ships until a human has signed off on all four licenses: code, weights, training data and dependencies.

Ten minutes from signup to first response.

Every step below is running today. Keys are issued from the console the moment you sign in.

  1. Create a key

    Sign up, name a project, reveal the secret once. Live and test keys are separate and revocable.

    pd_live_7e77c8deb5fc4b958c0…

  2. Hand us an image

    Send the bytes as multipart/form-data in an image part. JPEG, PNG or WebP, up to 10 MB and 50 megapixels once decoded.

    -F "image=@person.jpg"

  3. Read the JSON

    One synchronous call. Stable error codes, the request ID in every log line, and the exact model digest that produced the result.

    POST /v1/face/analyze → 200

The four-gate rights review

Vision APIs you're actually allowed to ship.

A model is more than its code. Weights, training data and bundled components each carry rights of their own, and any one of them can block commercial use. We treat missing, research-only and unclear as blocked, and the paperwork exists for every model you call.

  1. Model code license
  2. Model weight license
  3. Training-dataset rights inherited by the weights
  4. Bundled dependency and component licenses

Ask us for the rights dossier behind any service you call and we will send it.

Stateless by default

Neither face service creates an enrollment record or stores an embedding. Analyze images and results expire after 7 days; Detect retains no image or response geometry.

Your images stay yours

No customer image is ever used for training without a separate, explicit consent agreement. EXIF is stripped on ingest.

Apparent, not actual

Ancestry and gender outputs are visual classifications with calibrated confidence and an unknown outcome. Never proof of identity, and never for hiring, credit, housing or policing.

Reproducible releases

Every response names the model versions and SHA-256 weight digests behind it, so you can pin a version and diff a regression.

A clear rate for every service.

Pay for successful images, not a bundle of unrelated features. Live face API rates are shown per 1,000 images so they are easy to compare; unreleased services stay unpriced until their production benchmark is complete.

Provisional · pricing finalizes at GA

No checkout is live today. Face Analyze and Face Detect are currently free within account limits.

  • Face Analyze

    /v1/face/analyze

    Live
    Today
    Free during beta
    GA rate
    $1.20 / 1,000$0.0012 per image
    Billed unit
    Successful image
  • Face Detect

    /v1/face/detect

    Public beta
    Today
    Free during beta
    GA rate
    $0.80 / 1,000$0.0008 per image
    Billed unit
    Successful image
  • Background Removal

    /v1/image/matte

    License review
    Today
    Not available
    GA rate
    After benchmarkNo estimate published
    Billed unit
    Successful image
  • Restore & Upscale

    /v1/image/enhance

    License review
    Today
    Not available
    GA rate
    After benchmarkNo estimate published
    Billed unit
    Successful image
  • Image to Image

    /v1/image/transform

    License review
    Today
    Not available
    GA rate
    After benchmarkNo estimate published
    Billed unit
    Successful image
  • Face Recognition

    /v1/face/match

    Planned
    Today
    Not available
    GA rate
    After benchmarkNo estimate published
    Billed unit
    Match request

Paid plans will be sold by Paddle, our merchant of record. No charges are being collected today. Paddle checkout is not active yet.

Public price comparison

Face detection for less than the big clouds.

PixelDepot Face Detect is 20% below the comparable AWS entry rate and 47% below Google Cloud Vision. Face Analyze adds apparent attributes and still lists below Google face detection.

Public USD list prices checked August 7, 2026

  • 20% lowerFace Detect versus the AWS entry rate
  • 47% lowerFace Detect versus Google Cloud Vision
  • $0 todayBoth face APIs during beta, within account limits
PixelDepot face API prices compared with public competitor list prices
ProviderServiceUSD / 1,000Billing basisPrice context
PixelDepotFace Detect$0.80Successful images after the shared free allowance20% below AWS and 47% below Google
PixelDepotFace Analyze$1.20Successful images after the shared free allowanceApparent attributes included; 20% below Google
AWS RekognitionDetectFaces$1.00First 1 million Group 2 images each monthFalls to $0.80, $0.60 and $0.25 at higher tiers
Google Cloud VisionFace Detection$1.50Requests 1,001 through 5 million each monthFirst 1,000 free; $0.60 above 5 million
Luxand.CloudBundled Face API$0.50 to $6.00Effective rate across monthly request plansCan be lower at 200,000 or more committed requests

Rates are normalized to 1,000 requests. Free tiers, volume tiers, monthly commitments, storage, support and taxes differ. PixelDepot GA rates remain provisional. Company and product names belong to their respective owners; no affiliation or endorsement is implied.

  • Free allowance

    $0/ first 1,000 face images

    A monthly allowance for evaluating the API before paid usage begins at GA.

    • Shared across Face Detect + Analyze
    • Only successful images count
    • Test + live keys
    • Community support
    Create an account
  • Pay-as-you-go

    Most teams

    From $0.0008/ image (by service)

    After the free allowance. The service table above is the pricing source of truth.

    • $10 minimum top-up
    • Unlimited keys and projects
    • Every released service
    • Volume discounts from 250k images / month
    • Usage + latency dashboard
    • 99.9% availability target at GA
    Create an account
  • On-premises

    Custom/ annual

    The same Docker inference plane running inside your own network or venue.

    • Self-hosted GPU or CPU image
    • Air-gapped deployment
    • Model pinning + private builds
    • Named engineer at GA
    Create an account

Use case · migration

Replacing Kairos? Start with Face Analyze.

Face Analyze covers the attribute half of Enroll -- age, gender presentation, ancestry and glasses -- without creating a recognition record for someone who just walked past a photo booth. The geometry half is thinner: a box per face, and no landmarks, pose or quality score. The table below is the whole map.

Kairos Enroll

PixelDepot v1

topLeftX / topLeftY / w / h
bounding_box
age
attributes.estimated_age
gender.type + confidences
attributes.perceived_gender_presentation
asian / black / hispanic / white
attributes.apparent_ancestry
glasses
attributes.has_glasses
leftEyeCenter / rightEyeCenter
no equivalent
eyeDistance
no equivalent
yaw / pitch / roll
no equivalent
quality
no equivalent
lips
no equivalent

The gallery fields (face_id, subject_id, gallery_name) have no equivalent: Face Analyze returns the analysis without enrolling anyone. Errors arrive as HTTP status codes, not as 200 with an Errors array.

Questions we get asked

Before you put us in a proposal.

Can I call the API today?

Yes. Create an account, issue a key from the console, and call POST /v1/face/analyze, or the geometry-only /v1/face/detect beta. Every other service in the catalog carries its own status badge.

When do the other services arrive?

When each one passes its rights review and its evaluation, and not before. We would rather miss no date than miss a date we published. The catalog badges track where each service stands, and the beta list hears first when one opens.

Can I process EU images?

Geometry, pose and quality: yes. Apparent-attribute features, including apparent ancestry and apparent gender, are unavailable for EU processing, because the EU AI Act prohibits biometric categorization used to infer them. Where the server sits does not change that, so hosting in the US is not a workaround.

What about US biometric law?

Illinois BIPA and Texas CUBI govern collecting face geometry, and they apply to you as the operator of the booth or app. Neither face service creates an enrollment record or identity template, which is a much smaller footprint than recognition, but it does not remove your notice and consent obligations. This is a description of our product, not legal advice.

What happens to the images I send?

Face Analyze images and results are retained for 7 days so your account can show what came back, then deleted automatically. We monitor the oldest surviving payload so a stalled deletion job is visible. Face Detect retains neither the image nor response geometry and shows operational metadata only. No face embedding is stored, and no customer image trains a model without a separate written agreement.

Who am I actually buying from?

PixelDepot is operated by Foto Master Ltd, Ilan Ramon 5, Ness Ziona, Israel, and paid plans will be sold by Paddle.com Market Limited as merchant of record, the same arrangement as our other products. Paddle handles payment, tax and refunds. No charges are being collected today. Paddle checkout is not active yet.

Is there an SLA?

Not today, because there is nothing to apply one to. The Scale plan carries a 99.9% availability target at GA. Anything stronger is a contract conversation, not a page on a website.

What if a model gets pulled for license reasons?

It can happen, and it is the reason the four-gate review exists. Every response names the model version and the SHA-256 digest of the weights behind it, so you can pin a version and see exactly what changed. If a license position changes under a service you are calling, you hear it from us with the reason.

Can I run it myself?

That is the On-premises plan: the same Docker inference plane inside your own network or venue, including air-gapped deployment and pinned model versions. Tell us on the beta form and we will scope it with you.

Create an account.

Sign in with Google or a link to your inbox. No password. An account means a console and API keys that the public endpoints answer today.

Signing in creates the account if you do not have one.