BudgetPixel Depth Maps API

openapi: 3.2.0
info:
  contact:
    email: support@budgetpixel.com
    name: BudgetPixel Support
  description: 'The BudgetPixel developer API for programmatic image, video, music and

    sound-effect generation.


    Generation is **asynchronous**: you create a job, then poll its status until it

    reaches a terminal state (`succeeded` / `failed`). You are charged in credits

    only on success — never for failures, timeouts, or content blocked before

    generation.


    **Authentication.** All requests require an API key sent as a bearer token:


    Authorization: Bearer bpx_live_xxx


    Create and manage keys from your BudgetPixel account dashboard. The developer

    API is available on **every paid plan**. On the **Free** plan, API keys can call

    the free library searches only: `GET /v1/stock/images`,

    `GET /v1/stock/sound-effects` and `GET /v1/stock/music`.


    **Pricing.** API usage is metered in credits at each model''s published price

    (see `GET /v1/models`), and your plan''s model discounts and active promotions

    apply to API charges exactly as they do on the web — `POST /v1/cost` quotes

    your discounted price. Pro and Ultra plans include a daily free-models

    allowance (a shared pool of free generations per day across qualified image

    models such as FLUX 2 Klein and Qwen-Image) that covers API

    requests too — the standard price applies once the day''s pool is used.

    Other promotional free daily generations remain web/app-only. Failed

    generations are never charged.


    **Rate limits.** Requests are limited per minute: **600 requests/min per API

    key** and **1200 requests/min per source IP** (defaults; subject to tuning).

    Every response carries `X-RateLimit-Limit`, `X-RateLimit-Remaining`, and

    `X-RateLimit-Reset` (seconds until the window resets). Exceeding a limit

    returns HTTP `429` with a `Retry-After` header — wait that long before

    retrying. Rejected requests still count against the window, so a tight retry

    loop keeps you limited; back off instead. Generation jobs additionally have a

    per-plan concurrency cap (the same one as the web workshop), independent of

    these request limits. `POST /v1/uploads` also has a dedicated **60 uploads/min

    per account** cap (it''s free/unmetered) on top of these limits. Every `429`

    this API returns carries `Retry-After`, including a full concurrency queue

    (a suggested wait) and the monthly spend limit (seconds until the month

    resets).


    **Output retention.** Generated files are kept for 24 hours after the job is

    created. Job status responses give the exact time as `expires_at`; after it,

    `expired` is `true` and no download URLs are returned.'
  title: BudgetPixel Depth Maps API
  version: '2026-06-20'
servers:
- description: Production
  url: https://api.budgetpixel.com/v1
security:
- ApiKeyAuth: []
tags:
- description: Turn a video into a depth map video to guide other video models with its structure and motion
  name: Depth Maps
paths:
  /depth-maps/{id}:
    get:
      description: 'Poll until `status` is `succeeded`. The depth video is returned as a

        short-lived signed `video_url` (valid for several hours; re-call for a fresh

        URL anytime — the underlying object is deleted 24h after generation).


        The depth video is an H.264 MP4 at the input''s frame size (up to 1080p,

        or the `resolution` you asked for) and frame rate (up to 30 fps). In

        grayscale, near is white and far is black.


        With `face_mesh` or `include_face_landmarks`, the response adds

        `face_mesh` (how much of the clip had a face, and each face''s frames)

        and, with `include_face_landmarks`, a signed `face_landmarks_url`.'
      operationId: getDepthMapJob
      parameters:
      - description: The opaque job id returned by the create endpoint.
        in: path
        name: id
        required: true
        schema:
          type: string
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/DepthMapJob'
          description: Current job status. `video_url` is populated once succeeded.
        '401':
          $ref: '#/components/responses/Unauthorized'
        '403':
          $ref: '#/components/responses/Forbidden'
        '404':
          $ref: '#/components/responses/NotFound'
      summary: Get depth map job status
      tags:
      - Depth Maps
  /depth-maps/video-depth-anything:
    post:
      description: 'Turns a video into a depth map video: near is white, far is black, steady from frame to frame, at the input''s size (up to 1080p) and frame rate (up to 30 fps). Feed it to a video model as a structure or motion reference. Three sizes: small is fastest, base adds edge detail, large draws the finest detail (hair, thin poles, distant people) and is the steadiest. Priced per second of the INPUT video, rounded up, 5 seconds minimum: small 20, base 30, large 40 credits/second (input up to 30 seconds, ≤100MB). Output is H.264 MP4. Optional face mesh, at no extra charge: face_mesh draws each face''s mesh (eyes, brows and lips in colour) over the depth so a video model keeps lip, eye and head movement, and include_face_landmarks returns per-frame landmarks, expression scores and head pose as JSON.


        **Asynchronous.** Returns a job `id` (the video is not in this response). Poll `GET /v1/depth-maps/{id}` until `status` is `succeeded` — the depth video''s URL is in that response''s `video_url`.


        Examples, sample output and pricing: Video Depth Anything'
      operationId: createDepthMap_video_depth_anything
      requestBody:
        content:
          application/json:
            schema:
              properties:
                colormap:
                  default: grayscale
                  description: grayscale (near = white) is what video models take as a depth reference; turbo, inferno, magma and viridis colour it for viewing.
                  enum:
                  - grayscale
                  - turbo
                  - inferno
                  - magma
                  - viridis
                  type: string
                face_mesh:
                  default: false
                  description: Draw a mesh over each face (up to max_faces), with eyes, brows and lips in colour, so a video model given the depth map keeps lip, eye and head movement. Faces are found automatically; a face turned to full profile gets no mesh. No extra charge.
                  type: boolean
                include_face_landmarks:
                  default: false
                  description: 'Also return face_landmarks_url: a JSON file with 478 3D points, 52 expression (blendshape) scores and a head-pose matrix per face per frame. Works with or without face_mesh. The file is kept 7 days. No extra charge.'
                  type: boolean
                keep_audio:
                  default: false
                  description: Copy the input's audio into the depth video.
                  type: boolean
                max_faces:
                  default: 4
                  description: Most faces tracked at once (1-4), largest first. Applies to face_mesh and include_face_landmarks.
                  maximum: 4
                  minimum: 1
                  type: integer
                resolution:
                  default: auto
                  description: The output's short side; auto keeps the input's size up to 1080p. Never upscales. The price is the same at every resolution.
                  enum:
                  - auto
                  - 360p
                  - 480p
                  - 720p
                  - 1080p
                  type: string
                side_by_side:
                  default: false
                  description: Return the input on the left and the depth map on the right, twice the width.
                  type: boolean
                size:
                  default: small
                  description: 'Model size, which sets the per-second rate: small 20, base 30, large 40 credits.'
                  enum:
                  - small
                  - base
                  - large
                  type: string
                video:
                  description: The video to turn into a depth map (up to 30 seconds, MP4/MOV/WebM, ≤100MB). A public video URL or an uploaded-file URL from POST /v1/uploads (videos are passed by URL, not inlined).
                  type: string
                  x-bpx-media: video
              required:
              - video
              type: object
        required: true
      responses:
        '200':
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/CreateDepthMapResponse'
          description: Job accepted.
        '400':
          $ref: '#/components/responses/BadRequest'
        '401':
          $ref: '#/components/responses/Unauthorized'
        '403':
          $ref: '#/components/responses/Forbidden'
        '429':
          $ref: '#/components/responses/TooManyRequests'
      summary: Make a depth map with Video Depth Anything
      tags:
      - Depth Maps
components:
  schemas:
    CreateDepthMapResponse:
      properties:
        credits:
          charged when it succeeds.: null
          description: Total price of this job (seconds x the size's per-second rate). Held now
          type: integer
        id:
          description: Opaque job id (`dpt_...`) — use it to poll status.
          type: string
        message:
          type: string
        model:
          type: string
        seconds:
          description: Billed duration — the input video's length, rounded up to whole seconds, at least 5.
          type: integer
        size:
          description: The model size the job runs at.
          enum:
          - small
          - base
          - large
          type: string
        status:
          $ref: '#/components/schemas/JobStatus'
      type: object
    DepthMapJob:
      allOf:
      - $ref: '#/components/schemas/MotionJob'
      - properties:
          face_landmarks_url:
            description: 'With `include_face_landmarks`, once `status=succeeded`: a short-lived signed URL to a JSON

              file (`format: budgetpixel-face-landmarks/1`) with, per frame and per face, 478 `[x, y, z]`

              landmarks (x and y as fractions of the frame), 52 blendshape scores (names in

              `blendshape_names`) and a 4x4 head-pose `matrix`. Kept 7 days.

              '
            format: uri
            type: string
          face_mesh:
            description: Present when the job asked for `face_mesh` or `include_face_landmarks`, once the worker has run.
            properties:
              coverage:
                description: Share of frames with at least one face, 0 to 1. 0 means no face was found, so nothing was drawn.
                type: number
              drawn:
                description: 'Whether the mesh was drawn on the video (`face_mesh: true`).'
                type: boolean
              faces:
                description: One entry per tracked face.
                items:
                  properties:
                    first_frame:
                      type: integer
                    frames:
                      description: Frames this face has a mesh in
                      including bridged gaps.: null
                      type: integer
                    id:
                      type: integer
                    interpolated_frames:
                      description: Frames filled in across a short gap in tracking (up to 0.3 s).
                      type: integer
                    last_frame:
                      type: integer
                  type: object
                type: array
              frames:
                description: Frames in the output.
                type: integer
              frames_with_face:
                type: integer
            type: object
        type: object
    QueueFull:
      description: '`429` body when your account already has as many jobs of this kind

        (image, video or audio) in flight as its concurrency cap allows. Submit

        the next job when one of your running jobs finishes.

        '
      properties:
        active_jobs:
          description: Jobs of this kind in flight right now.
          type: integer
        code:
          enum:
          - queue_full
          type: string
        error:
          description: Human-readable message.
          type: string
        queue_full:
          enum:
          - true
          type: boolean
        queue_limit:
          description: Your concurrency cap for this kind of job.
          type: integer
        retry_after_seconds:
          description: Same value as the `Retry-After` header.
          type: integer
      required:
      - error
      - code
      - queue_full
      - queue_limit
      - active_jobs
      - retry_after_seconds
      type: object
    Error:
      properties:
        error:
          properties:
            code:
              description: Stable machine-readable code.
              example: model_not_available
              type: string
            message:
              type: string
            type:
              description: Error category.
              example: invalid_request_error
              type: string
          required:
          - type
          - code
          - message
          type: object
      required:
      - error
      type: object
    ModerationBlocked:
      description: Returned (with HTTP 400) when the input content moderation gate blocks a generation request. The block is a property of the request's prompt or input media — reword the prompt or change the input and retry. Branch on `restriction_reason`, which is stable and machine-readable.
      properties:
        error:
          description: Human-readable explanation of the block.
          type: string
        restriction_reason:
          description: Stable machine-readable block reason.
          enum:
          - input_csam
          - input_explicit_adult
          - input_upload_nudity
          - input_celebrity_likeness
          - strict_model_nsfw
          type: string
      required:
      - error
      - restriction_reason
      type: object
    JobStatus:
      description: Lifecycle state. `succeeded`/`failed`/`timeout` are terminal.
      enum:
      - pending
      - starting
      - processing
      - completing
      - succeeded
      - failed
      - timeout
      type: string
    MotionJob:
      properties:
        completed_at:
          format: date-time
          type: string
        created_at:
          format: date-time
          type: string
        error_message:
          type: string
        id:
          type: string
        model:
          type: string
        status:
          description: Lifecycle state. `succeeded`/`failed`/`canceled` are terminal.
          enum:
          - pending
          - processing
          - succeeded
          - failed
          - canceled
          type: string
        video_codec:
          description: Codec of the MP4 at `video_url`; present once `status=succeeded`. Video upscales are `h265` (HEVC, 10-bit); motion-control outputs are `h264`.
          enum:
          - h264
          - h265
          type: string
        video_codec_note:
          description: Present only when `video_codec` is `h265` — a short explanation of where H.265 plays and what to do when it does not.
          type: string
        video_url:
          description: Short-lived presigned URL; present once `status=succeeded`.
          format: uri
          type: string
      type: object
  responses:
    TooManyRequests:
      content:
        application/json:
          schema:
            oneOf:
            - $ref: '#/components/schemas/Error'
            - $ref: '#/components/schemas/QueueFull'
      description: 'A limit was reached. Every `429` carries `Retry-After` (seconds). The

        per-minute rate limit, the upload limit and the hourly API post cap answer

        with the `Error` envelope, code `rate_limited`. The monthly spend limit answers with

        code `monthly_spend_limit_reached`, and its `Retry-After` counts down to the

        start of next month (raising the limit in the console clears it sooner).

        A full concurrency queue answers with the flat `QueueFull` body (3D jobs:

        the `Error` envelope with code `queue_full`); its `Retry-After` is a

        suggested wait, because a slot frees as soon as one of your running jobs

        finishes.

        '
      headers:
        Retry-After:
          description: Seconds to wait before retrying.
          schema:
            minimum: 1
            type: integer
    NotFound:
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
      description: The requested job was not found.
    Unauthorized:
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
      description: Missing or invalid API key.
    BadRequest:
      content:
        application/json:
          schema:
            oneOf:
            - $ref: '#/components/schemas/Error'
            - $ref: '#/components/schemas/ModerationBlocked'
      description: The request was malformed, referenced an unavailable model, or was blocked by input content moderation (moderation blocks carry a `restriction_reason` — see ModerationBlocked).
    Forbidden:
      content:
        application/json:
          schema:
            $ref: '#/components/schemas/Error'
      description: Authenticated but not permitted (plan gate, ownership, account restriction). Content-moderation blocks are 400, not 403.
  securitySchemes:
    ApiKeyAuth:
      bearerFormat: bpx_live_*
      description: 'API key as a bearer token: Authorization: Bearer bpx_live_xxx'
      scheme: bearer
      type: http