How to Use Seed Audio 1.0 API for AI Voice Generation

PiAPI PiAPI July 10, 2026

Seed Audio 1.0 API through PiAPI lets you create an asynchronous speech-generation task with model: byteaudio and task_type: seed-audio-1.0, then poll the task until it returns an audio URL. It supports text input, optional voice references, output controls, and per-second billing for AI voice generation workflows.

This guide walks through the practical API flow: what you need before you start, how to create a task, how to poll the result, how to use voice references, and how to estimate pricing. If you want to test the workflow before writing code, open the Seed Audio 1.0 API playground and try a script in PiAPI Signal Studio first.

In this guide

To use Seed Audio 1.0 API through PiAPI, create a task at POST https://api.piapi.ai/api/v1/task with your X-API-Key, model: byteaudio, task_type: seed-audio-1.0, and a text input. PiAPI returns a task ID. Poll GET /api/v1/task/{task_id} until the task is completed, then use output.audio_url to play or download the generated audio.

Seed Audio 1.0 through PiAPI is an asynchronous BytePlus speech-generation API for turning text into spoken audio. Developers create a PiAPI task with model: byteaudio and task_type: seed-audio-1.0, then poll the task until PiAPI returns an audio URL.

The important word is speech. PiAPI's current Seed Audio 1.0 workflow is for spoken audio from text, with optional reference audio or image guidance. If your project needs music generation, use ACE-Step; if you need video-matched sound effects or action-timed audio, use the MMAudio API instead of treating Seed Audio 1.0 as a full sound-scene generator.

These are the core Seed Audio 1.0 API facts to keep nearby while building. API details were checked against the PiAPI Seed Audio documentation on July 10, 2026.

1. A PiAPI account and API key.
2. A short script to synthesize.
3. Output settings, such as format and sample rate.
4. Optional reference audio or image guidance if you want the result to follow a specific voice or style.

Seed Audio 1.0 through PiAPI is not a streaming API. You submit a task, wait for it to process, then retrieve the generated audio URL when the task is complete.

The core request uses PiAPI's standard task endpoint. This is the only full cURL example in the guide, so treat it as the base pattern.

curl --request POST \ --url https://api.piapi.ai/api/v1/task \ --header "Content-Type: application/json" \ --header "X-API-Key: YOUR_API_KEY" \ --data '{ "model": "byteaudio", "task_type": "seed-audio-1.0", "input": { "text": "Hello from PiAPI. This is a Seed Audio 1.0 test.", "format": "mp3", "sample_rate": 24000 } }'

Keep your API key on the server side. Do not expose it in frontend code, public repositories, browser console snippets, or client-side apps.

Seed Audio 1.0 API tasks are asynchronous. Instead of receiving the final audio immediately, your app should poll the get-task endpoint until the task reaches a terminal status.

1. Submit the task.
2. Store the returned task ID.
3. Poll GET /api/v1/task/{task_id} every few seconds.
4. Continue while the task is staged, pending, or processing.
5. Stop when the task is completed or failed.
6. On success, read output.audio_url and output.duration.

You do not need a different API request for every use case. In most cases, the important part is the script, the selected settings, and whether you add references.

Seed Audio 1.0 can use reference inputs when you want the generated speech to follow a voice or style instead of relying on a default voice. According to PiAPI's Seed Audio documentation, one request can include up to three audio references.

• Use up to three audio references per request.
• Assign them with @Audio1, @Audio2, and @Audio3.
• Use one image reference only when using image guidance.
• Do not mix audio and image references in one request.
• Keep reference files up to 10 MB.
• Keep URL audio references public HTTPS and 30 seconds or shorter.

PiAPI lists Seed Audio 1.0 pricing by successful output duration: $0.002 per successful output second, which is approximately $0.12 per generated minute.

For the latest product details, check the Seed Audio 1.0 API pricing and limits section on the product page before launching production usage.

Seed Audio 1.0 can behave like a text-to-speech API for simple scripts, but PiAPI exposes more controls than a basic preset-voice TTS workflow. If you are comparing reference-audio speech models, keep F5-TTS API open as a related option.

• Reference audio: up to three audio references.
• Multi-speaker script markers: @Audio1, @Audio2, and @Audio3.
• Image guidance: one image reference, separate from audio references.
• Output controls: format, sample rate, speech rate, pitch, and loudness.
• Processing model: async task submission and polling.

If you are still testing scripts, voices, or formats, you do not need to start with code. Open the Seed Audio 1.0 generator in PiAPI Signal Studio, paste one of the example scripts above, choose the output settings, and listen to the result.

• The task is asynchronous, not streaming.
• One request can generate up to 120 seconds of audio.
• Text input is limited to 2,048 Unicode characters.
• Returned audio URLs are temporary, so save results you need to keep.
• Audio references and image references cannot be mixed in the same request.
• URL audio references must be public HTTPS files and no more than 30 seconds.
• PiAPI's current Seed Audio endpoint should not be described as a music, ambience, or full sound-scene generator unless that behavior is separately documented and tested.

Most failed tests come from input shape, reference handling, account state, or output handling rather than the script idea itself.

• Empty or failed task: confirm input.text is present and moderation-safe.
• Reference ignored: use the documented input.references shape and public HTTPS URLs.
• Upload rejected: keep reference files under 10 MB and use supported formats.
• Audio URL expired: download or persist the result soon after completion.
• Unexpected voice: match @AudioN markers to reference order.
• Insufficient credits: add credits or shorten the test output.

If you still cannot identify the issue, contact the PiAPI team at support@piapi.ai with the task ID, request time, error message, and a short description of what you were trying to generate. Do not send API keys or private reference files over email.

Now you know how to use Seed Audio 1.0 API through PiAPI: create a task, poll the result, download the audio, and add reference voices when the workflow needs more control.

For a fast first pass, test your script in the Seed Audio 1.0 API playground . For production integration, use the PiAPI Seed Audio documentation to confirm the latest request fields, limits, and response shape.

If you want evidence before choosing a script, you can also hear Seed Audio 1.0 handle difficult scripts across names, numbers, punctuation, long narration, and one transparently reported reference-audio failure.

Disclaimer: Use reference voices only when you have the rights or consent needed to use that voice, and review the applicable PiAPI, BytePlus, content, and jurisdictional terms before publishing or monetizing generated audio.