Skip to documentation

Connect only what you can operate.

Audoryx normalizes provider behavior without hiding ownership, cost, capability, quota, or legal differences. Credentials are supplied and controlled by the installation operator.

Provider security model

Open Admin → Settings → API Settings. Each provider drawer shows only the connection inputs needed for that provider.

  • Saved secrets are encrypted under the application key.
  • The browser receives connection state and a short masked hint, never the decrypted credential.
  • Leaving a configured secret blank retains its saved value.
  • Connection tests execute server-side and return sanitized results.
  • Operational endpoints, timeouts, formats, and internal model defaults are not exposed as routine administrator inputs.
  • Disable a provider before revoking its external account credential.
BYOK is not a grant of resale rights

Confirm API resale/OEM permissions, end-user access, voice consent, retention, content rules, commercial output ownership, and regional availability directly with every provider.

Capability matrix

ProviderConnectionPrimary useImportant boundary
ElevenLabs MusicAPI keyComplete music generation and detailed music responses where the account/API supports them.Do not present unrelated speech products as Audoryx music features.
Google LyriaAI Studio key or Vertex service-account/OAuth connectionLyria 2 instrumental output and supported Lyria 3 preview song workflows.Model availability, region, preview terms, output format, and duration differ.
MusicGPTOfficial MusicGPT API keyTwo-version song generation, Pro-model HLS streaming, Voice to Song remixing, extend, inpaint, stems, WAV, and timed lyrics.Streaming is available on v6-pro and v7-pro for paid MusicGPT accounts; Audoryx copies completed files immediately.
Suno-compatible APIAPI key for the configured independent compatible serviceAsynchronous generation, supported advanced workflows, uploaded-vocal Voice to Song, and compatible editing actions.Audoryx is not affiliated with Suno; the operator must validate the chosen service and contract.

ElevenLabs Music

Create a permitted provider credential

Use an account and API plan whose terms cover the intended production and end-user workflow.

Paste only the API key

Open the ElevenLabs drawer, enter the key, enable the connection when ready, and save. Base URLs, output formats, default models, and stream timeouts are server-managed.

Test the connection

The real health check verifies account identity and usable access. A saved key without a passing test is not production-ready.

Run one authorized generation

Confirm early metadata, stream availability when returned safely, final audio, duration, lyrics/timing behavior, cover assignment, credit settlement, and stored media.

Timestamped lyric alignment may use supported internal ElevenLabs capabilities. Audoryx does not expose a general transcription, narration, TTS, dubbing, isolation, sound-effects, or stem product through this music connection.

Google Lyria

Choose the connection path that matches the Google account and model:

AI Studio

Enter the AI Studio API key when the desired Google music model and account support that connection path.

Vertex AI

Provide the required project identity and authorized service-account JSON or OAuth connection. Grant only the least privileges required by the selected model.

Model familyAudoryx treatment
Lyria 2Instrumental music generation. It does not accept lyrics or generate vocals.
Lyria 3 previewSupported preview song/clip workflows with behavior defined by the current Google preview offering and connected account.

Review the official Google documentation for the current model name, region, quota, preview status, duration, format, and terms. Audoryx's connection test does not replace a permitted end-to-end production test.

MusicGPT

The official adapter exposes MusicGPT v6, v6-pro, v7, and v7-pro through Music AI v2. The -pro models deliver an HLS URL for paid API accounts by token-bound webhook while each variation is still being composed; Audoryx relays the update through its authenticated generation stream and loads browser HLS support only for those live sources.

  • Configure the MusicGPT API key and select API Plus or API Pro in Admin → API Settings; unavailable models and streaming capabilities are then disabled automatically.
  • MusicGPT songs receive provider-owned Remix / Cover, Extend, Replace Section (Inpaint), Stem Separation, and WAV actions.
  • Voice to Song, Remix, Extend, and Inpaint use MusicGPT's provider-managed v6 workflows; v7 is never advertised for an endpoint that cannot accept it.
  • The source provider, task, conversion, and media IDs are always derived from the completed owned song and cannot be replaced by browser input.
  • Each completed MP3, WAV, and stem is copied into private Audoryx storage immediately.
  • The connection test validates the public HTTPS callback, then uses the official read-only voice catalogue without submitting a billable generation.
  • The finance ledger applies plan-aware per-song and duration-based rates for MusicGPT generation and studio tools.

Suno-compatible provider

This adapter targets an independently operated, configurable compatible API. The default service identity in a build does not create an official relationship with Suno.

  • Enter the API key in its connection drawer and run the real health test.
  • Ensure the installation's public HTTPS callback routes are reachable by the configured service.
  • Simple and Advanced generation, Audio, Voice, Inspo, Voice to Song, remix, extension, replacement, stems, and export controls appear only where the selected model and operation support them.
  • Voice to Song uploads a user-owned vocal performance and builds an arrangement around it; it is not text-to-speech.
  • Do not change provider endpoints in production without validating authentication, response schemas, callback signatures/tokens, idempotency, SSRF boundaries, media limits, and legal terms.
UI capability controls are enforced twice

The studio hides or disables incompatible controls for clarity, while the server independently rejects unsupported operations. Client state is never the authorization boundary.

Models and costs

Open Admin → AI models & costs. Every visible model should have an accurate name, provider, capability set, enabled state, artwork, user credit multiplier, and provider cost basis.

  • Disable a model until the connected account can use it successfully.
  • Set credit cost from measured provider cost and the intended margin, not from a guessed label.
  • Document whether a result produces one or multiple tracks.
  • Verify Simple, Advanced, Voice to Song, lyrics, vocals, streaming, and editing support separately.
  • Retest after a provider model, API version, pricing, region, or account entitlement changes.

Streaming generation lifecycle

StateUser experienceOperator expectation
QueuedThe request is accepted locally and waiting for a worker.A worker must consume the correct queue promptly.
SubmittedEarly provider metadata may appear before final audio.The provider accepted or is processing the task.
StreamingPlay becomes available only after a safe playable stream URL exists.Do not mark a failed or metadata-only item as Live.
CompletedFinal media, duration, cover, lyrics/timing, actions, and credit settlement are available.Persist and verify provider media before exposing owner-authorized playback.
FailedA sanitized message is shown; no Live line or label remains.Release/refund reserved credits exactly once and retain an actionable server-side diagnostic.

Production verification

  1. Use a provider test account whose terms permit the intended workflow.
  2. Run the connection test from the final production origin.
  3. Generate every advertised operation at least once.
  4. Verify queue, callback/polling, live updates, stream playback, final media, lyrics, artwork, costs, credits, and failures.
  5. Review sanitized logs for secrets, raw provider payloads, personal audio, or unsupported claims.
  6. Record the provider account, tested model, date, result, and contract owner in the private release evidence.

Deeper adapter behavior remains documented in the provider engineering reference.

Verified release walkthrough

Visual guide

These annotated captures come from the reviewed release listed in the screenshot manifest. Follow the numbered steps below each image.

Audoryx AI models and costs page with configured-model summary, creator availability toggles, credit multipliers, base credits, provider costs, and save actions.
AI model availability and cost controls
  1. 1
    Model summary Confirm how many models are configured and enabled for creators before changing individual entries.
  2. 2
    Creator availability Enable a model only after its provider connection and cost basis have been verified.
  3. 3
    Credit calculation Base credits and the multiplier determine the visible credit charge for a creator request.
  4. 4
    Provider cost basis Store the provider-side cost and billing unit used for internal finance and profitability reporting.
Audoryx creator workspace showing early Suno song titles and artwork, Streaming now badges, enabled play controls, and active playback before generation completes.
Suno playback while generation is still streaming
  1. 1
    Early playable outputs Metadata, artwork, and audio controls appear as soon as the provider exposes a streamable result.
  2. 2
    Streaming state The status remains explicit while the full composition continues processing in the background.
  3. 3
    Playback before completion The persistent player confirms that the stream can be heard without waiting for the terminal provider callback.
Audoryx administrator provider settings with connected Suno and ElevenLabs cards, disabled Google Lyria, capability summaries, and connection checks.
Provider connection and capability routing
  1. 1
    Connection-only configuration Administrators manage provider credentials and availability without exposing internal URL, timeout, or output-format tuning.
  2. 2
    Visible capability map Each card explains which production capabilities and models become available when the connection is active.
Audoryx documentation · Version 2.0.0 · Updated 3 September 2026