generated: '2026-07-19'
method: searched
source: https://docs.ltx.io/authentication, https://docs.ltx.io/async-jobs, https://docs.ltx.io/errors,
  https://docs.ltx.io/rate-limits, https://docs.ltx.io/debugging, openapi/lightricks-ltx-openapi.yml
api: LTX API
base_url: https://api.ltx.video
authentication:
  style: bearer-token
  header: Authorization
  format: 'Bearer YOUR_API_KEY'
  issued_at: https://console.ltx.io/
  env_var_convention: LTXV_API_KEY
  oauth: false
  docs: https://docs.ltx.io/authentication
  artifact: authentication/lightricks-authentication.yml
idempotency:
  supported: false
  notes: >-
    LTX documents no idempotency key. Generation is inherently non-idempotent (each submit bills and
    produces a new job id). Callers must dedupe client-side by correlating their own request key with
    the returned job `id` before retrying a submit.
async_jobs:
  supported: true
  style: submit-and-poll
  submit: POST /v2/{endpoint}
  submit_status: 202
  submit_response_fields: [id, created_at]
  poll: GET /v2/{endpoint}/{id}
  poll_interval_seconds: 5
  status_values: [pending, processing, completed, failed]
  terminal_states: [completed, failed]
  result_field: result
  result_shape: map of output label to signed URL (e.g. video_url, exr_frames_url)
  retention_hours: 24
  retention_notes: After 24 hours in a terminal state, GET /v2/{endpoint}/{id} returns 404.
  endpoints: [text-to-video, image-to-video, audio-to-video, retake, extend, video-to-video-hdr,
    video-to-video-reframe]
  docs: https://docs.ltx.io/async-jobs
sync_mode:
  supported: true
  version: v1
  notes: v1 endpoints return the rendered media directly as application/octet-stream. Long renders can
    hit connection timeouts; v2 async is the documented recommendation for production.
pagination:
  supported: false
  notes: No list/collection endpoints — the API is job-oriented, not resource-collection-oriented.
field_expansion:
  supported: false
metadata:
  supported: false
request_tracing:
  response_header: x-request-id
  description: Unique identifier for each API request. Include it when contacting support.
  recommendation: Log request IDs in production deployments.
  docs: https://docs.ltx.io/debugging
versioning:
  scheme: uri-path
  versions: [v1, v2]
  current: v2
  notes: v1 is the synchronous surface and v2 the asynchronous job surface — they are parallel modes,
    not a strict version succession. Model versioning is separate and carried in the `model` request
    field (ltx-2-3-fast / ltx-2-3-pro).
  artifact: lifecycle/lightricks-lifecycle.yml
error_envelope:
  format: custom
  shape: '{"type": "error", "error": {"type": "<error_type>", "message": "<human readable>"}}'
  rfc9457: false
  discriminator: error.type
  artifacts: [errors/lightricks-error-codes.yml, errors/lightricks-problem-types.yml]
  docs: https://docs.ltx.io/errors
rate_limit_signaling:
  status: 429
  retry_after_header: true
  limit_headers_documented: false
  limit_types: [concurrency_limit_error, rate_limit_error]
  default_concurrency: 2
  artifact: rate-limits/lightricks-rate-limits.yml
  docs: https://docs.ltx.io/rate-limits
file_inputs:
  upload_endpoint: POST /v1/upload
  flow: >-
    create-upload returns a pre-signed upload_url (PUT, expires in 1 hour) plus a storage_uri to
    reference in image_uri / video_uri parameters. Uploaded files are available for 24 hours.
  public_url_alternative: true
  docs: https://docs.ltx.io/input-formats
billing:
  model: per-second-of-media
  credits: true
  auto_top_up: https://docs.ltx.io/auto-top-up
  docs: https://docs.ltx.io/pricing