AudioShake

AudioShake takes audio that already exists and pulls it apart: a song into vocals, drums, bass and other stems, a finished film or podcast mix into dialogue, music and effects, or a conversation into one track per speaker. It does not create new music. The choice is mostly about how you will run it. Studio is the browser workspace and is the sensible place to try a few files. The API suits a pipeline, but its price is set in credits per minute per model, and the pages we read do not give a dollar price per credit.

Free Studio gives 200 credits, MP3 export and files up to 10 minutes. WAV and FLAC export appear from Starter at $20 a month, so a producer who needs lossless stems should plan for that tier. Editorial assessment: use Free to judge separation quality on one or two of your own tracks before paying, and ask the provider for the API credit price before any batch job.

Credits are charged per minute of source audio, for every model you request on that file, and each file’s length is rounded up to the next minute. The table applies the published rates to a few jobs; the credit counts are arithmetic from the models page, not a quote in dollars.

The 10 credits given at API sign-up therefore cover roughly one 10-minute stem job, or one minute of multi-speaker separation. Failed tasks that stop before processing starts are not charged, according to the API FAQ. The pages do not explain how Studio’s monthly credits map to these API rates, so do not assume the same numbers apply.

Separating a recording does not change who owns it. If you plan to release a remix, an instrumental or a clean dialogue track made from someone else’s audio, settle the licence for the source first. On the provider’s side, its terms, effective October 1, 2026, say customers keep their rights in their content and that AudioShake will not use that content to train AI models that generate new music. Those terms are written as a services agreement with order forms, so confirm with the provider that they cover a self-serve Studio account.

Also plan for short retention on the API: uploaded assets expire after 72 hours and download links after one hour, so save results as soon as a task finishes. Limits to know before you queue work: files up to 2 GB, multi-speaker input up to 1.5 hours, lyric work up to 45 minutes, and 60 API requests per second.

1. Separate it into the stems you actually need, and listen to each one against the original for bleed, missing parts and artefacts.
2. For video work, split a clip with dialogue and music and check whether the dialogue stem is clean enough to send to your captioning tool.
3. Note the file length limit and export format you hit on Free, and decide whether WAV or FLAC export or longer files would justify Starter.
4. If you plan to use the API, ask the provider for the dollar price per credit and price your own monthly minutes before building on it.

Leave the observed results blank until the run is done. Move to a paid plan only if the stems are usable for your task and you need the export or the credits.