generated: '2026-08-13'
method: searched
source: >-
  https://www.lalal.ai/api/v1/docs/ (HTTP 200), https://www.lalal.ai/pricing/,
  https://www.lalal.ai/business-solutions/, https://www.lalal.ai/llms.txt
provider: LALAL.AI
provider_id: lalal-ai

summary: >-
  LALAL.AI ships an interactive API CONSOLE but NO sandbox. There is no test mode, no test/live key
  separation, no magic test identifiers, no fixture or trigger tooling and no time simulation. Every call
  made from the console is a real call against production that consumes real processing minutes from the
  license key you paste in. What stands in for a sandbox is the free tier and the enterprise trial —
  real minutes, not fake ones.

console:
  present: true
  url: https://www.lalal.ai/api/v1/docs/
  http_status: 200
  title: LALAL.AI Public API v1
  engine: Scalar API Reference
  engine_evidence: >-
    The page loads https://s.lalal.ai/less/pages/scalar/scalar.css and declares
    <link rel="alternate" type="application/json" href="/api/v1/openapi.json" title="OpenAPI JSON">.
  spec_source: https://www.lalal.ai/api/v1/openapi.json
  interactive: true
  sends_live_requests: true
  credentials_required: true
  credential: X-License-Key (a real license key from a Pro plan or a minute top-up)
  note: >-
    Rendered client-side from the OpenAPI document, so the console and the machine-readable contract can
    never drift apart. But it is a live-fire console: there is no test key that would let you exercise an
    operation without spending minutes.

test_vs_live:
  separated: false
  key_prefixes: none
  modes: none
  note: >-
    One credential type exists — the account license key in the X-License-Key header. It is not prefixed,
    not scoped, not rotatable through the API, and there is no test counterpart. The same key that runs a
    console experiment runs production.

magic_test_values:
  present: false
  note: >-
    No test file ids, no reserved source_ids, no simulated failure triggers. The only way to exercise an
    error path is to cause it for real — e.g. reuse a spent idempotency_key to observe
    `idempotency_key_used`, or exhaust minutes to observe `exceeded_duration`.

time_simulation:
  present: false
  note: >-
    No test clocks. The two time-dependent behaviours — the 24-hour source-file retention and the monthly
    Fast-minute renewal — can only be observed in real time.

fixtures_and_triggers:
  present: false

free_evaluation:
  - id: no-registration-preview
    name: Unregistered web preview
    url: https://www.lalal.ai/
    api_access: false
    description: >-
      Verbatim from the provider's llms.txt — "Most tools can be tried right away, no registration
      required: upload a file, get a free preview of the result, and listen to it online before signing up
      or paying." Web product only; it does not issue an API license key.
  - id: starter
    name: Starter (free plan)
    url: https://www.lalal.ai/pricing/
    api_access: false
    quota: 10 Relaxed-queue minutes, lifetime
    restriction: Online listening only — no downloads, no batch, no API.
    description: >-
      The free plan cannot be used to evaluate the API. API access requires a Pro plan or a minute top-up
      pack.
  - id: enterprise-trial
    name: Business Solutions free trial
    url: https://www.lalal.ai/business-solutions/
    request_url: https://www.lalal.ai/business-solutions-request/
    api_access: true
    quota: 30 minutes
    gated: sales-form
    description: >-
      The only route to unpaid API minutes: a free 30-minute trial offered on the Business Solutions page,
      obtained by submitting the enterprise quote/trial form. Not self-serve.
  - id: paid-entry
    name: Minute top-up (no subscription)
    url: https://www.lalal.ai/upgrade/
    api_access: true
    description: >-
      The self-serve route to API access without a subscription — buy a top-up pack (750 / 3,000 / 5,000
      Fast minutes). Minutes never expire.

minute_consumption:
  formula: file_length x number_of_selected_stems
  note: >-
    Worth knowing before running console experiments: a multistem call selecting six stems bills six times
    the file duration. There is no dry-run that returns the cost without doing the work. Call
    POST /api/v1/limits/minutes_left/ before and after to measure it.

cross_references:
  - plans/lalal-ai-plans-pricing.yml
  - rate-limits/lalal-ai-rate-limits.yml
  - authentication/lalal-ai-authentication.yml