Hermoso AI Generations API

openapi: 3.2.0
info:
  title: Hermoso Ai Generations API
  version: '2026-08-26'
  description: 'Operations tagged generations across 2 of this provider''s published API definitions: app_hermoso_ai.json, app_hermoso_ai_rest.json. Each path carries the servers of the definition it was published in.'
servers:
- url: https://app.hermoso.ai
security:
- bearerAuth: []
tags:
- name: Generations
paths:
  /v1/generations/quote:
    post:
      operationId: quoteGeneration
      summary: Price a generation before running it
      description: 'The estimated credits a generation with this body would reserve, from the render’s own pricing, with the brand’s product photos attached exactly as the real run attaches them. Free: nothing is queued, reserved or charged. It is an estimate: the render is charged its actual cost when it finishes, usually within a few percent of this figure. `credits: null` with `refused` means the render would not start, and says why.'
      tags:
      - Generations
      parameters:
      - name: X-Hermoso-User
        in: header
        required: false
        schema:
          type: string
        description: 'Which profile to act on, when the account holds more than one: an `id` from `GET /v1/brands`. Defaults to the workspace the key is pinned to. Membership is re-authorised on every request, and an id this key cannot act on is a 404 `brand_not_found`, never an empty result.'
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/GenerationCreate'
      responses:
        '200':
          description: Success
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/GenerationQuote'
        '400':
          description: Invalid request.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
        '401':
          description: Missing, invalid or revoked key — or a channel that is not connected, in which case `error.connector` names it and `error.type` is `connector_error`.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
        '404':
          description: '`error.code: brand_not_found`: `X-Hermoso-User` (or the profile this key is pinned to) names a profile this key cannot act on. `GET /v1/brands` lists the valid ids.'
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
        '429':
          description: Rate limited. `Retry-After` says how long to wait.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
      security:
      - bearerAuth: []
    servers:
    - url: https://app.hermoso.ai
  /v1/generations:
    post:
      operationId: createGeneration
      summary: Generate an image or video
      description: 'Queue one image or one video render for this profile and answer 202 with the job. Poll `GET /v1/jobs/{id}` until `status` is `done` (its `url` is the file) or `error`. Billed exactly as the same render is anywhere in Hermoso: the credits are reserved when it starts and settle to the measured cost. A refusal (not enough credits, a prompt the model will not render) is an error in the standard envelope, with the server’s own sentence in `message`. Send an `Idempotency-Key` and a retry never starts a second render.'
      tags:
      - Generations
      parameters:
      - name: callback_url
        in: query
        required: false
        schema:
          type: string
        description: 'The same as the `X-Hermoso-Callback-Url` header, for a client that cannot set headers. Prefer the header: a callback URL is a secret, and query strings are logged.'
      - name: X-Hermoso-Callback-Url
        in: header
        required: false
        schema:
          type: string
          format: uri
        description: 'An https:// URL to POST the outcome to, for a platform that cannot poll (for example Zapier `z.generateCallbackUrl()`). When the call queued a job, Hermoso POSTs the job resource — exactly what `GET /v1/jobs/{id}` answers — once the job is `done` or `error`, with headers `Hermoso-Event: job.done | job.error` and `Hermoso-Delivery` (also sent as `Idempotency-Key`; drop a repeat). A 2xx ends delivery; a 4xx other than 408, 425 and 429 is final; anything else is retried for about 16 hours. When the call finished inside the request, its response body is POSTed too (`Hermoso-Event: tool.result`). The response header `Hermoso-Callback` says what happened: `registered`, `sent`, `limit-reached` or `not-registered`. Nothing about what runs or what it costs changes: the URL is attached after the job is queued. A refusal is the HTTP error and nothing is POSTed. Must be https on a public host name; a replay with the same `Idempotency-Key` and a new URL adds that URL to the same job.'
      - name: Idempotency-Key
        in: header
        required: false
        schema:
          type: string
        description: 'Send a unique value to make this call safe to retry. A replay returns the original response with `Idempotency-Replayed: true`, for 24 hours — except a call that queued a job, whose replay is that job as it is now (what `GET /v1/jobs/{id}` answers), so a finished render replays as done with its file.'
      - name: X-Hermoso-User
        in: header
        required: false
        schema:
          type: string
        description: 'Which profile to act on, when the account holds more than one: an `id` from `GET /v1/brands`. Defaults to the workspace the key is pinned to. Membership is re-authorised on every request, and an id this key cannot act on is a 404 `brand_not_found`, never an empty result.'
      requestBody:
        required: true
        content:
          application/json:
            schema:
              $ref: '#/components/schemas/GenerationCreate'
      responses:
        '200':
          description: Success
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Job'
        '202':
          description: Queued. Poll `poll` (also sent as `Location`).
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Job'
        '400':
          description: Invalid request.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
        '401':
          description: Missing, invalid or revoked key — or a channel that is not connected, in which case `error.connector` names it and `error.type` is `connector_error`.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
        '404':
          description: '`error.code: brand_not_found`: `X-Hermoso-User` (or the profile this key is pinned to) names a profile this key cannot act on. `GET /v1/brands` lists the valid ids.'
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
        '429':
          description: Rate limited. `Retry-After` says how long to wait.
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/Error'
      security:
      - bearerAuth: []
    servers:
    - url: https://app.hermoso.ai
components:
  schemas:
    GenerationQuote:
      type: object
      properties:
        object:
          type: string
          enum:
          - quote
        type:
          type: string
        credits:
          type: number
          nullable: true
          description: What the render reserves. null when it would not start (see `refused`).
        likely_credits:
          type: number
          nullable: true
          description: 'Present when the hold covers an optional step (a label repair) that usually does not run: the likely charge, `credits` being the most it can be.'
        label:
          type: string
          nullable: true
          description: The model or pipeline that will render it.
        refused:
          type: string
          nullable: true
          description: Why the render would not start. Nothing was charged.
        reference_images:
          type: array
          items:
            type: string
          description: 'Image only: the pictures the render will use, exactly as it attaches them. With `use_brand` on these are the profile’s own saved product photos (up to three) and its logo when the prompt asks for one; empty when none is on file.'
        reference_note:
          type: string
          nullable: true
          description: 'Image only: a sentence about those pictures when there is one to say (a photo left out, a lineup, the logo).'
    Job:
      type: object
      properties:
        object:
          type: string
          enum:
          - job
        id:
          type: string
        type:
          type: string
          description: 'What kind of work it is: video, stitch, avatar, templatead, upscale, and so on. This set MAY GROW.'
        status:
          type: string
          description: 'queued, running, done or error. `done` and `error` are final. This enum MAY GROW: keep polling on a value you have not seen.'
        progress:
          type: number
          description: 0 to 1, when the worker reports it.
        label:
          type: string
          nullable: true
        brand_id:
          type: string
          description: The profile that started the job.
        created_at:
          type: string
          nullable: true
          description: ISO-8601.
        updated_at:
          type: string
          nullable: true
          description: ISO-8601. For a finished job, when it finished.
        run_at:
          type: string
          nullable: true
          description: ISO-8601, for work scheduled to start later; null for work that starts immediately.
        url:
          type: string
          nullable: true
          description: The finished file, once `status` is `done` and the job produced one. Always an absolute, publicly reachable URL.
        poster_url:
          type: string
          nullable: true
          description: A still for a finished video, when there is one.
        model:
          type: string
          nullable: true
          description: The model that rendered it.
        credits_used:
          type: number
          nullable: true
        error_message:
          type: string
          nullable: true
          description: Why it failed, when `status` is `error`. Named `error_message` so it can never be mistaken for this API’s error envelope.
        cancelled:
          type: boolean
          description: True when the job was stopped by `POST /v1/jobs/{id}/cancel` before it reached a model. Its `status` is `error` and nothing was charged.
        result:
          type: object
          nullable: true
          description: 'The worker’s full result. PASSTHROUGH TIER: its shape is the job type’s own and improves with it, like a tool’s `data`. Build on the fields above; read this for detail.'
        poll:
          type: string
          description: This job’s own path.
    GenerationCreate:
      type: object
      required:
      - type
      - prompt
      properties:
        type:
          type: string
          enum:
          - image
          - video
          description: What to make. This enum MAY GROW.
        prompt:
          type: string
          description: What the ad should show, in plain words.
        model:
          type: string
          description: A model id from the live catalog (the `hermoso_capabilities` tool lists them with prices). Omit to let Hermoso pick; the quote then prices the model the server would pick.
        aspect_ratio:
          type: string
          description: e.g. 1:1, 4:5, 9:16, 16:9. Each model lists the ratios it renders.
        image_size:
          type: string
          description: 'Image only: a size tier the model lists (for example 1K, 2K).'
        duration_seconds:
          type: integer
          description: 'Video only: one of the chosen model’s listed lengths.'
        ref_image:
          type: string
          description: 'Video only: a Hermoso-hosted image to animate (image-to-video).'
        ref_images:
          type: array
          items:
            type: string
          description: 'Image only: up to 3 Hermoso-hosted images (a `POST /v1/media` or Library `url`) the render uses as its reference. When present the brand’s product photos are not attached for you; the quote prices exactly this.'
        use_brand:
          type: boolean
          description: 'Image only, default true: attach the profile’s own product photos (and its logo when the prompt asks for one) so the render shows the real product. The quote prices exactly what this attaches.'
        format:
          type: string
          enum:
          - png
          - gif
          - svg
          description: 'Image only, default png. gif = a short looping animated GIF, svg = an editable vector. Both render on gpt-image-2.6 (default) or gpt-image-2.6-flare from the prompt alone: no ref_images, and the profile’s photos are not attached.'
        frames:
          type: integer
          minimum: 2
          maximum: 16
          description: 'format gif only: frames in the loop (default 9).'
        frame_ms:
          type: integer
          minimum: 20
          maximum: 2000
          description: 'format gif only: milliseconds per frame, rounded to 10 (default 200).'
        complexity:
          type: string
          enum:
          - low
          - medium
          - high
          description: 'gpt-image-2.6 models only: how much the model plans before drawing (default high). No separate charge.'
    Error:
      type: object
      required:
      - error
      properties:
        error:
          type: object
          required:
          - type
          - code
          - message
          properties:
            type:
              type: string
              enum:
              - authentication_error
              - permission_error
              - invalid_request_error
              - connector_error
              - rate_limit_error
              - insufficient_credits
              - api_error
              description: The coarse family — branch on this. New CODES land inside an existing type wherever possible so that branch keeps working; the enum is published for code generation, and a client should still fall through gracefully on a member it does not recognise.
            code:
              type: string
              description: The specific reason. Log this. New values may appear at any time.
            message:
              type: string
            param:
              type: string
              description: The offending field, when there is one.
            connector:
              type: string
              description: Present when the reason is a channel that is not connected. Reconnect in the app — an OAuth consent screen cannot be completed headlessly.
            credits_needed:
              type: number
              description: 'On `insufficient_credits` from a recast or an edit: what the request needs.'
            balance:
              type: number
              description: 'On `insufficient_credits` from a recast or an edit: what this workspace has.'
            unknown_arguments:
              type: array
              items:
                type: object
                properties:
                  name:
                    type: string
                    description: The key that was sent.
                  did_you_mean:
                    type: string
                    nullable: true
                    description: An argument THIS tool declares that the key most likely meant, or null.
              description: 'On `invalid_arguments` from POST /v1/tools/{name}: each key the tool does not take, with the argument it most likely meant. Nothing was run.'
            upgrade_url:
              type: string
              description: 'On `real_face_needs_paid`: where to choose a paid plan.'
            alternative:
              type: string
              description: 'On `real_face_needs_paid`: the field that works on every plan instead (`person.describe`).'
            person_url:
              type: string
              description: 'On a recast with `person.describe` whose recast step was refused after the person was made: the AI person, in your Library. Send it as `person.image_url` to retry without making a new one.'
        request_id:
          type: string
          nullable: true
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      bearerFormat: hmk_...
      description: A Hermoso API key, or an access token from the OAuth authorization-code flow (which is the same thing).
x-refined-from:
- app_hermoso_ai.json
- app_hermoso_ai_rest.json