Sources & references
Stable Audio product overview ↗
Create music or sound effects from a prompt, then work with the result in a track-based studio. Stable Audio charges credits when it records new audio; mixing, tape edits and export are free operations on existing material.
Stable Audio is Stability AI’s music and sound creation workspace. Its current guide describes a browser studio built around a session and a recording deck rather than a single download button. A prompt can create a full mix on one track or an experimental multitrack arrangement with separately editable parts. The latter lets a producer keep a useful bass line while replacing an unsuitable drum part, but the arrangement agent may interpret the prompt less literally. This makes the tool especially interesting for iterative sound design.
Describe genre, instruments, mood, tempo and target duration before starting. The guide offers fixed durations of 30 seconds, one minute, 90 seconds and three minutes, with an automatic option; the broader product page also discusses longer creations. For a first test, make a 30-second loop and judge it inside the actual video, game scene or performance. If one bar is weak, use replace-section or regenerate-track rather than repeatedly creating a new complete song. Keep an ear on the transition around the repaired region, because a technically smooth splice can still feel musically abrupt.
The guide supplies unusually clear credit mechanics: a full-mix start costs four credits, an experimental multitrack start six, and adding, regenerating, replacing or extending audio four each. Playback, faders, pan, mute, solo, effects, tape edits, bouncing, export and sharing do not create new audio and are listed as free. The public pricing page is rendered dynamically and does not expose a reliable subscription amount in the text available for this review. Treat the credit schedule as a way to estimate effort, then confirm the current plan price and commercial license in the signed-in account.
Short film cue
Game ambience
Songwriting sketch
Official product image. Click to inspect the details.

Use full mix when a single finished cue is enough; choose experimental multitrack when separate drums, bass or keys will need later adjustment.
Replace one unsatisfactory part without discarding the rest of a useful arrangement. Compare the new part at the boundaries and in the full mix.
Select a time range to re-record a bad passage, or extend a track beyond its current end. Both operations create audio and consume credits.
Adjust volume, pan, mute, solo, reverse, stutter or freeze existing material without an additional generation charge. Changes should serve the scene rather than decorate it.
Ask the producer panel about a session or request an operation; if it records new audio, that action costs the same credits as its manual equivalent. Export after listening to the final mix.
Write a musical brief
Generate one controlled take
Repair selectively
Mix, license-check and export
Specify instrumentation, pace, mood, duration and where the piece will be heard; check that any reference material is yours to provide.
Start with a short full mix, or multitrack when you need separate parts, and note the initial credit cost.
Solo tracks, identify a weak instrument or time range, and regenerate only that portion instead of discarding the whole session.
Set levels, check loop boundaries and commercial terms, then export a version tested alongside the final visual or interactive scene.
See plan details below.
The current Stable Audio guide counts credits only for actions that record new audio: four for a full-mix session, six for a multitrack session, and four for each additional track, regeneration, section replacement or extension. It states that failed generations are not charged. Existing audio may be mixed, edited, bounced, played and exported without further generation credits.
Stable Audio offers a way to start, but the public subscription-price page is client-rendered and did not provide a dependable numerical plan comparison for this review. The cost of a month therefore depends on your credit package, how often you regenerate, and whether the intended use is permitted under the applicable license. Confirm those terms in the current checkout; do not infer that a free creation automatically carries the same commercial rights as a paid plan.
No. The guide charges for newly recorded audio; playback, mixing, tape edits, effects and export are listed as free.
The current guide lists four credits.
Starting an experimental multitrack session costs six credits, regardless of how many initial tracks it returns.
Yes. Regenerate a track or replace a selected time range while keeping other material.
The current web-studio guide says up to eight; consolidate or remove tracks to make room.
Do not assume so; confirm the plan and current license for the precise distribution use before publishing.
Explore platforms, inputs and outputs, licensing, and access requirements.
The web studio guide describes up to eight active tracks per session. Consolidating tracks frees space but commits levels, pan and effects; the guide describes only a limited one-hour undo window. A tempo-and-key attachment for multitrack mode is used to infer those musical properties, not to copy the attached audio into the result. The separate DAW plugin and API should be evaluated against their own documentation.
Prompts and uploaded reference material can reveal unreleased creative work. Only attach audio you have permission to submit, especially when matching tempo and key. If sharing a session link, check who can access it and avoid treating a private work-in-progress as public marketing material. Review the current Stability AI terms and privacy policy for training, retention and licensing before customer work.
Stable Audio product overview ↗
Loading experiences…
Bring a real task and see how it fits the way you work.
User experiences