Music Models

Stability AI

Stable Audio 2.5

Music and sound via Replicate, with an explicit caveat about the audited app's duration parameter.

Timing caveatReplicate

At a glance

Plan the music before you generate.

Provider reference & app audit
Gateway · owner audit
Replicate
App status · owner audit
Listed
App request
60 seconds · no duration selector
Duration forwarded
seconds_total

Official provider / gateway limit

Replicate's published schema accepts duration from 1 to 190 seconds, default 190.

Read the official references below

The audit reports seconds_total = 60 (wrapper cap 190). Replicate documents duration instead. Acceptance and actual output length are unverified; this is not a guarantee of 60-second output.

Reported upstream model: stability-ai/stable-audio-2.5

App picker ID: stability-ai/stable-audio-2.5

Capabilities & boundaries

What it supports. What the app exposes.

Official documentation

  • Replicate describes text-to-audio generation for music and sounds.
  • The published API schema includes duration, seed, diffusion steps and guidance controls. These are provider settings, not controls exposed by the audited app.
Sources for these capabilities

Unverified behavior & limitations

  • Important schema conflict: the audit reports seconds_total, while current Replicate documentation uses duration. This website does not repair or validate that application request.
  • The README mentions continuation from an audio clip, but the inspected schema does not establish that workflow for this app route. No audio-upload control is exposed here.
  • Generated quality, exact timing and prompt adherence were not tested. Use only material you have rights to use; review the applicable provider and gateway terms before publishing.

Owner audit · not a live test

App controls, upstream parameters and fallback behavior are reported by the owner-supplied October 9, 2026 code audit, not independently inspected application code or live generation tests. Official documentation verifies provider capabilities, not successful execution in EZ Ai Assist.

Controls in EZ Ai Assist

  • Main prompt: 1–500 characters, according to the owner audit. There is no separate lyrics input or vocals toggle for this model.
  • The app requests durationSeconds: 60. There is no user-selectable duration control.

Upstream parameters · reported by audit

  • prompt and seconds_total = 60, according to the owner audit. The reported wrapper caps this value at 190.
  • No seed, steps or guidance override is reported.

Put it to work

Start with a more useful musical brief.

Original examples from EZ Ai Assist, not generated or listening-tested samples. Review the integration caveats above before using them.

Idea 01

Stable Audio 2.5: First musical direction

Original EZ Ai Assist example for the main prompt in standard mode where applicable. Instructions describe a creative target, not guaranteed output or an added control.

Make an instrumental lo-fi study bed with a soft electric piano phrase, subdued bass and gentle brushed percussion. Keep the energy stable and the arrangement uncluttered, with subtle harmonic variation and no vocals, loud fills or sudden endings.

Idea 02

Stable Audio 2.5: A contrasting arrangement

Original EZ Ai Assist example for the main prompt in standard mode where applicable. Instructions describe a creative target, not guaranteed output or an added control.

Create a minimalist electronic instrumental for a calm software walkthrough. Use a light sequenced pulse, warm sustained pads and occasional bell accents. Avoid dominant lead melodies so narration remains clear; no singing or spoken words.

Idea 03

Stable Audio 2.5: Another creative brief

Original EZ Ai Assist example for the main prompt in standard mode where applicable. Instructions describe a creative target, not guaranteed output or an added control.

Compose a quiet acoustic instrumental sketch with nylon-string guitar and a small warm bass part. Use a relaxed swaying feel, leave space between phrases and keep the texture intimate. Avoid crowd noise, vocals and heavily processed effects.

Developer reference

Provider pricing is not your app charge.

Provider or gateway reference pricing is not EZ Ai Assist pricing. Picker estimates and Premium labels are not verified provider costs or actual customer charges. Check the app's current charge before generating.

A model-specific gateway rate is not established in this guide. No price is inferred from a picker label, another gateway or a provider subscription. See the official references for current terms.

Public reference rates can change and do not establish the actual cost of a request. No paid request or billing record was used for this review.

View EZ Ai Assist plans

Common questions

Stable Audio 2.5 FAQs

Can I choose the track length in EZ Ai Assist?

The app requests durationSeconds: 60. There is no user-selectable duration control. The audit reports seconds_total = 60 (wrapper cap 190). Replicate documents duration instead. Acceptance and actual output length are unverified; this is not a guarantee of 60-second output.

What is the provider's maximum?

Replicate's published schema accepts duration from 1 to 190 seconds, default 190.

Which inputs does the app expose?

Main prompt: 1–500 characters, according to the owner audit. There is no separate lyrics input or vocals toggle for this model. The app requests durationSeconds: 60. There is no user-selectable duration control.

Is the picker price what I will be charged?

Provider or gateway reference pricing is not EZ Ai Assist pricing. Picker estimates and Premium labels are not verified provider costs or actual customer charges. Check the app's current charge before generating.

Has this app integration been independently tested?

App controls, upstream parameters and fallback behavior are reported by the owner-supplied October 9, 2026 code audit, not independently inspected application code or live generation tests. Official documentation verifies provider capabilities, not successful execution in EZ Ai Assist. Important schema conflict: the audit reports seconds_total, while current Replicate documentation uses duration. This website does not repair or validate that application request.

Sources & freshness

Check the original references.

Reviewed . These are official provider or gateway references; a gateway is not the same as the model developer. New and Legacy describe the owner-audited app listing, not release dates or retirement guarantees.

Same provider