generated: '2026-07-19'
method: derived
source: openapi/hedra-web-api-openapi-original.json
entities:
- name: Asset
  description: An uploaded or generated media object (image, audio, video, 3D).
  id_field: id
  key_fields: [id, type, name, created_at, asset, thumbnail_url]
  types: [image, audio, video, 3d]
- name: Generation
  description: An asynchronous generation job that produces an Asset.
  id_field: id
  key_fields: [id, type, input, status, progress, created_at, credit_cost, asset, error, batch_generation_id]
  statuses: [queued, processing, finalizing, complete, error]
- name: AIModel
  description: A generation model (Hedra Character-3/Avatar/Omnia or partner models like Veo, Kling, Grok, Flux, Sana).
  id_field: slug
  key_fields: [slug, name, type, price_details, resolutions, durations, inputs]
- name: Voice
  description: A voice usable for text-to-speech or voice-clone.
  id_field: external_id
  key_fields: [type, external_id, labels, preview_url, source]
- name: CreditBalance
  description: The account/workspace billing credit balance.
  key_fields: [WorkspaceCreditUsage]
relationships:
- from: Generation
  to: Asset
  type: has_one
  via: asset
  note: A completed generation produces one (or batch of) asset(s); asset_id on GenerationStatusResponse.
- from: Generation
  to: AIModel
  type: belongs_to
  via: model_slug
  note: Each generation selects a model by model_slug (ai_model_id deprecated).
- from: Generation
  to: Asset
  type: has_many
  via: start_keyframe_id / end_keyframe_id / audio_id / video_id / reference_image_ids / reference_video_ids / reference_audio_ids
  note: Video/avatar generations reference input assets by id.
- from: Generation
  to: Voice
  type: belongs_to
  via: voice_id
  note: Text-to-speech generations select a voice by voice_id.
- from: Generation
  to: Generation
  type: has_many
  via: batch_generation_id
  note: Batch generations (batch_size 1-8) are linked by a shared batch_generation_id.
id_prefixes:
  note: Ids are opaque; models are addressed by stable `slug`, generations/assets by server-issued `id` (UUID-like). Clients may pre-reserve ids via reserved_asset_id / generation_id.