import requests

response = requests.post(
    f"https://api.stability.ai/v2beta/audio/stable-audio-2/audio-to-audio",
    headers={"authorization": f"Bearer sk-MYAPIKEY", "accept": "audio/*"},
    files={
        "audio": open("./uk-bass-base.mp3", "rb"),
    },
    data={
        "prompt": "A song in the 3/4 time signature that features cheerful acoustic guitar, live recorded drums, and rhythmic claps, The mood is happy and up-lifting.",
        "output_format": "mp3",
        "duration": 20,
        "model": "stable-audio-2.5",
    },
)

if response.status_code == 200:
    with open("./output.mp3", "wb") as file:
        file.write(response.content)
else:
    raise Exception(str(response.json()))

"The bytes of the generated mpeg"

Stable Audio 2

Audio-to-Audio

Stable Audio transforms existing audio samples into new high-quality compositions up to three minutes long at 44.1kHz stereo using text instructions. Discover techniques for sample transformation in our Audio to Audio Guide to maximize creative control. Read more about the model capabilities here.

Try it out

Grab your API key and head over to or try Stable Audio 2.0 for free at stableaudio.com.

How to use

Please invoke this endpoint with a POST request.

The headers of the request must include an API key in the authorization field. The body of the request must be multipart/form-data. The accept header should be set to one of the following:

audio/* to receive the audio in the format specified by the output_format parameter.
application/json to receive the audio encoded as base64 in a JSON response.

The body of the request should include:

prompt - text to generate the audio from. Check our prompt guide for tips
audio - the audio to use as the starting point for the generation

Notes:

We do not allow copyrighted content to be uploaded to our platform.

Maximum request size is 50Mb.

Optional Parameters:

The body may optionally include:

output_format - the format of the output audio
seed - the randomness seed to use for the generation
steps - the number of sampling steps
duration - the number of seconds of the generated audio
cfg_scale - controls how strictly the diffusion process adheres to the prompt text (only for stable-audio-2)
model - the model to use [stable-audio-2, stable-audio-2.5]
strength - controls how much influence the audio parameter has on the output audio

Note: for more details about these parameters please see the request schema below.

Credits

Stable Audio 2.0

By default, 20 credits per successful generation. The number of credits is determined by the following formula: credits = 17 + 0.06 * steps. Examples:

50 steps = 20 credits [default]
100 steps = 23 credits

Stable Audio 2.5

Requests made using the Stable Audio 2.5 model have a flat rate of 20 credits per successful result.

As always, you will not be charged for failed generations.

POST

v2beta

audio

stable-audio-2

audio-to-audio

import requests

response = requests.post(
    f"https://api.stability.ai/v2beta/audio/stable-audio-2/audio-to-audio",
    headers={"authorization": f"Bearer sk-MYAPIKEY", "accept": "audio/*"},
    files={
        "audio": open("./uk-bass-base.mp3", "rb"),
    },
    data={
        "prompt": "A song in the 3/4 time signature that features cheerful acoustic guitar, live recorded drums, and rhythmic claps, The mood is happy and up-lifting.",
        "output_format": "mp3",
        "duration": 20,
        "model": "stable-audio-2.5",
    },
)

if response.status_code == 200:
    with open("./output.mp3", "wb") as file:
        file.write(response.content)
else:
    raise Exception(str(response.json()))

"The bytes of the generated mpeg"

Authorizations

authorization

string

header

required

Use your Stability API key to authentication requests to this App.

Headers

authorization

string

required

Your Stability API key, used to authenticate your requests. Although you may have multiple keys in your account, you should use the same key for all requests to this API.

Minimum string length: 1

content-type

string

required

The content type of the request body. Do not manually specify this header; your HTTP client library will automatically include the appropriate boundary parameter.

Minimum string length: 1

Example:

"multipart/form-data"

enum<string>

default:audio/*

Specify audio/* to receive the bytes of the audio directly. Otherwise specify application/json to receive the audio as base64 encoded JSON.

Available options:

audio/*,

application/json

stability-client-id

string

The name of your application, used to help us communicate app-specific debugging or moderation issues to you.

Maximum string length: 256

Example:

"my-awesome-app"

stability-client-user-id

string

A unique identifier for your end user. Used to help us communicate user-specific debugging or moderation issues to you. Feel free to obfuscate this value to protect user privacy.

Maximum string length: 256

Example:

"DiscordUser#9999"

stability-client-version

string

The version of your application, used to help us communicate version-specific debugging or moderation issues to you.

Maximum string length: 256

Example:

"1.2.1"

Body

multipart/form-data

prompt

string

required

What you wish the output audio to be. A strong, descriptive prompt that clearly defines instruments, moods, styles, and genre will lead to better results.

You can make a prompt as simple or complex as you like. Simple prompts are good for clean output audio. Complex prompts are good for adding texture and depth to the output audio.

Check our prompt guide for tips.

Maximum string length: 10000

audio

file

required

The audio to be use as the starting point for the generation.

Supported Formats:

Validation Rule:

Audio must be between 6 and 190 seconds long

duration

number

default:190

Controls the duration in seconds of the generated audio.

Required range: 1 <= x <= 190

seed

number

default:0

A specific value that is used to guide the 'randomness' of the generation. (Omit this parameter or pass 0 to use a random seed.)

Required range: 0 <= x <= 4294967294

steps

integer

Controls the number of sampling steps.

For stable-audio-2: accepts steps between 30 and 100 (defaults to 50).
For stable-audio-2.5: accepts steps between 4 and 8 (defaults to 8).

cfg_scale

number

How strictly the diffusion process adheres to the prompt text (higher values make your audio closer to your prompt).

Defaults to 7 for stable-audio-2 and 1 for stable-audio-2.5 if not specified.

Required range: 1 <= x <= 25

model

enum<string>

default:stable-audio-2

The model to use for generation.

stable-audio-2.5 requires 20 credits per generation
stable-audio-2 requires 20 credits per generation

Available options:

stable-audio-2.5,

stable-audio-2

output_format

enum<string>

default:mp3

Dictates the content-type of the generated audio.

Available options:

mp3,

wav

strength

number

default:1

Sometimes referred to as denoising, this parameter controls how much influence the audio parameter has on the generated audio. A value of 0 would yield audio that is identical to the input. A value of 1 would be as if you passed in no audio at all.

Minimum value for stable-audio-2.5 is 0.01.

Required range: 0 <= x <= 1

Response

Generation was successful.

The bytes of the generated audio.

The finish-reason and seed will be present as headers.