SeedAudio.co

SeedAudio.co offers prompt-led audio scenes and an asynchronous API, with reference inputs and duration-based credit billing.

SeedAudio.co offers a browser workspace for creating audio scenes with dialogue, background music and sound effects. The operator markets its generator as Seed Audio 1.0 and describes that model as ByteDance Seed technology. That description should not be taken as proof that ByteDance operates or endorses this particular website.

Describe the whole sound scene

The SeedAudio.co workspace asks for a scene prompt covering speakers, language, emotion and sound events. Its examples combine dialogue with ambience and music, while another example requests foley without speech. This makes the intended task broader than reading a script aloud: you describe the sound surrounding the speakers as well.

The site advertises audio scenes up to two minutes and optional references. Its API documentation allows up to three public audio URLs or one public image URL; the image cannot be combined with audio references in the same request. Keep a reference brief focused and use only material you have permission to upload or make publicly accessible.

Submit a task, then collect the result

The API documentation describes asynchronous generation. A request returns a task ID, which an application uses to poll for completion. Successful responses include an output URL and the generated duration. The listed output formats are MP3, WAV, PCM and Ogg Opus, with configurable sample rate and voice-related controls.

For an integration, preserve the task ID and handle failed or canceled states before offering a download. The documented idempotency key lets an identical repeated request return the original task without another charge. Reusing that key for a different request produces a conflict instead.

Budget for the reservation, not only the clip

API access requires a paid plan or credit purchase. Web and API requests share a credit balance. The documentation lists billing at 75 credits per generated minute, with a ten-credit minimum and a temporary 150-credit reservation per task. It says unused reserved credits return after success and failed tasks receive a full refund.

A short draft is a sensible place to assess dialogue clarity and the balance of music against speech before spending on longer scenes. Confirm current credit terms and permitted use before building a client project around the service; a model name and a downloadable file do not establish commercial rights.

Before you try it

Frequently asked questions

What inputs can guide a SeedAudio.co sound scene?

The web workspace accepts a prompt with optional reference audio or one reference image. It describes scenes containing dialogue, music, ambience and sound effects rather than only reading a supplied script aloud.

Official source
Does the SeedAudio.co API return completed audio immediately?

No. The API is asynchronous: submit a request, retain the returned task ID and poll its status until success or failure. A successful submission is therefore different from a finished audio file.

Official source
Why does a SeedAudio task reserve more credits than its final charge?

The API documentation says a task reserves 150 credits, covering its maximum 120-second duration. Final billing uses actual output length at 75 credits per minute, with a 10-credit minimum; unused reserved credits are returned after success and failed tasks are refunded.

Official source
Can I use the SeedAudio API with only free signup credits?

The API guide explicitly requires a paid plan or credit purchase. The Free pricing card also displays an API-access label, so its wording is inconsistent; use the documented authentication and billing requirements when planning an integration.

Official source