Prompt-driven sound scenes
Describe speakers, emotion, timing, ambience, background music, and sound events in prompts up to 2,048 characters.
Seed Audio 1.0 API
Seed Audio 1.0 API for production audio generation
Generate dialogue, ambience, BGM, Foley, and sound effects from one prompt. The Seed Audio 1.0 API gives your product a stable task interface, status polling, managed output storage, and predictable credit settlement.
API access is available after a plan or credit purchase.
POST /audio/generations202 Accepted
curl -X POST https://seedaudio.co/api/v1/audio/generations \
-H "Authorization: Bearer $SEED_AUDIO_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: request-001" \
-d '{
"model": "seed-audio-1.0",
"prompt": "A suspense radio drama in English"
}'Model
seed-audio-1.0
Mode
Asynchronous
Billing
75 credits/min
Minimum
10 credits
Formats
MP3 · WAV · PCM · OGG
Prompt limit
2,048 chars
One Seed Audio 1.0 API, fewer integration problems
Keep one public contract while provider selection, compatible fallback, output storage, and credit settlement happen behind the scenes.
One public contract
Integrate one versioned endpoint and one response format instead of maintaining separate provider clients.
Managed provider fallback
Evolink is the primary source, with compatible fallback handled internally when a provider is unavailable.
Predictable credit billing
Reserve credits at submission, settle against actual output duration, and automatically return the unused difference.
Seed Audio 1.0 API capabilities
Control the audio scene, references, and output
The API exposes the controls needed for product integration without leaking internal provider-specific fields into your application.
Audio or image references
Guide a task with public audio references or one reference image. A preset voice counts toward the limit of three total audio inputs, and audio cannot be combined with an image.
Voice and delivery controls
Select a supported preset voice and adjust sample rate, speed, volume, pitch, and output format when the workflow needs them.
Production-friendly output
Request MP3, WAV, PCM, or OGG Opus. Successful tasks return a stored audio URL, duration, format, and final credit usage.
Queryable task status
Create a task with POST, then poll its ID with GET until it becomes succeeded, failed, or canceled.
Safe request retries
Send an Idempotency-Key so a network retry returns the original task instead of generating and charging twice.
What teams build with the Seed Audio 1.0 API
Use the same asynchronous workflow for single assets, creator tools, or larger audio generation pipelines.
Video and short-film sound
Generate dialogue, room tone, transitions, background music direction, and Foley concepts for video scenes and prototypes.
Podcasts and audio drama
Create narrated scenes, multi-character story moments, mood beds, and episodic audio drafts from structured prompts.
Games and interactive content
Produce environmental ambience, character moments, UI sounds, and event-driven audio variations for rapid iteration.
BGM and creative direction
Turn descriptions of mood, pacing, instrumentation, and narrative function into background music drafts for a wider workflow.
Sound effects and Foley
Describe physical actions, materials, distance, space, and timing to generate sound-effect and Foley-style results.
Batch content pipelines
Use task IDs and idempotent creation to coordinate many independent audio jobs from your own queue or application backend.
From API key to audio URL
The API is asynchronous by design, so long-running generations never need to keep a request connection open.
1
Create an API key
Sign up, purchase credits, and create a key from your account settings.
2
Submit a generation
Send a prompt and optional audio or image references to the generation endpoint.
3
Poll for the result
Query the task ID every few seconds until the output URL and final usage are available.
Seed Audio 1.0 API pricing
Pay for actual generated duration
API tasks use the same credit wallet as the web product. Final usage is calculated at 75 credits per generated minute, rounded up to a whole credit, with a 10-credit minimum.
- • 150 credits are reserved before a task starts, preventing concurrent requests from overspending the wallet.
- • After success, the unused difference between the reservation and actual duration is returned automatically.
- • Failed tasks return the full reservation, and internal provider fallback does not create an extra user charge.
Example final charges
8s10 credits
30s38 credits
60s75 credits
120s150 credits
A good fit for asynchronous generation
Choose the Seed Audio 1.0 API when your product can create a job and wait for a generated asset rather than maintaining a live audio connection.
- • Backend services, creator tools, and queued generation workflows
- • Audio scenes that combine speech, ambience, music direction, and effects
- • Applications that need durable result URLs and usage records
Know the current API boundaries
The current version focuses on reliable task creation and polling. Designing around these boundaries will make the first integration smoother.
- • It is asynchronous and is not a real-time streaming speech API
- • Public user webhooks are not available yet; poll the task status endpoint
- • API access requires a paid plan or credit purchase and at least 150 available credits per new task
Seed Audio 1.0 API questions
Practical answers for evaluating and integrating the API.
What is the Seed Audio 1.0 API?+
It is an asynchronous audio generation API for creating dialogue, ambience, background music direction, Foley, and sound effects from prompts with optional reference media.
How do I call the Seed Audio 1.0 API?+
Create an API key, send a Bearer-authenticated POST request to /api/v1/audio/generations, save the returned task ID, and poll GET /api/v1/audio/generations/{id} every few seconds.
How is Seed Audio 1.0 API usage charged?+
The final charge is 75 credits per generated minute with a 10-credit minimum. A new task reserves 150 credits, then automatically returns the unused difference after completion.
Can I send reference audio or an image?+
Yes. Use public audio reference URLs or one public image URL. A preset voice counts toward the limit of three total audio inputs, and audio references cannot be combined with an image in the same task.
Do I need to integrate Evolink or Fal separately?+
No. Your application integrates only the Seed Audio 1.0 API contract. Provider selection and compatible fallback remain internal and do not change the public response format.
Does the API support real-time streaming or webhooks?+
Not in the current public version. Generation is asynchronous and clients query task status by ID. User-configurable webhooks can be added in a later version.
Ready to add AI audio to your product?
Create an account, purchase credits, and make your first Seed Audio API request in minutes.