Text-directed audio scenes
Describe speaker roles, emotional pacing, sound design, and ending cues in one prompt for Seed Audio 1.0.
Seed Audio 1.0 on FAL
Seed Audio 1.0 Generator for Text, Voice, and Image-Guided Audio
Build Seed Audio 1.0 requests with text, reference audio, or an image, then keep provider calls behind a signed-in Miso One workspace with server-side controls.
FAL model: bytedance/seed-audio-1.0$0.075 per generated minute on FALPrompt limit: 2048 charactersReference audio: up to 3 files
Interactive FAL UI
Generate with Seed Audio 1.0
Write a prompt, add optional reference context, and inspect a validated payload for the FAL bytedance/seed-audio-1.0 endpoint. Live provider calls stay paused until credits, entitlement, and rate limits are configured.
Prompt examples
bytedance/seed-audio-1.0
{
"prompt": "Create a 20-second launch narration for an AI voice product. Use a confident host, soft electronic bed, and a clean ending sting.",
"output_format": "mp3",
"sample_rate": 24000,
"speed": 1,
"volume": 1,
"pitch": 0
}2048
Maximum prompt characters accepted by the FAL schema
3
Reference audio URLs supported in one request
6
Sample-rate choices from 8 kHz to 48 kHz
4
Output formats: wav, mp3, pcm, and ogg_opus
What Is Seed Audio 1.0?
Seed Audio 1.0 is a ByteDance text-to-audio model available through FAL. It can turn written direction into natural audio while accepting optional reference audio or one image as guidance.
Seed Audio 1.0 moves beyond a narrow text to speech box. Instead of only reading a sentence in one voice, it can take a production-style prompt that describes speaker tone, ambience, music, timing, and scene intent.
The FAL API exposes the model as bytedance/seed-audio-1.0. The required prompt field can include text to synthesize and markers such as @Audio1, @Audio2, and @Audio3 when reference audio URLs are provided.
For reference-guided work, Seed Audio 1.0 accepts up to three short audio files, or a single image URL when you want visual context. Audio references and image references are separate modes and should not be sent together.
Miso One wraps those details in a crawlable model page and a safer request UI: users can inspect output_format, sample_rate, speed, volume, and pitch before a signed-in, configured workspace sends anything to FAL.
Seed Audio 1.0 Workflows
Local case images show practical Seed Audio 1.0 workflows for text-directed scenes, reference audio, and image-guided audio design.
Reference audio control
Pass audio_urls and use @Audio markers so the prompt can refer to up to three short clips by order.
Image-guided sound design
Use image_url when a visual frame should guide the tone, environment, or audio scene direction.
Text to natural audio
Write a detailed prompt and let Seed Audio 1.0 generate natural-sounding audio from the full direction, not only from a plain sentence.
Reference audio by order
Attach up to three audio URLs and point to them inside the prompt with @Audio1, @Audio2, and @Audio3 markers.
Image-guided generation
Send a single jpeg, png, or webp image URL when the audio should follow a visual frame or scene concept.
Format and sample-rate control
Choose mp3, wav, pcm, or ogg_opus and set sample_rate to match your editing or delivery pipeline.
Safer server-side API key flow
The browser sends only validated generation settings. The FAL key stays on the server through fal_api_key or FAL_KEY configuration once live generation is enabled.
SEO-ready model context
The page adds crawlable FAQs, HowTo data, model facts, and internal links for people comparing AI audio generation tools.
How to Use Seed Audio 1.0
The first viewport UI mirrors the FAL SeedAudioInput schema while keeping validation, billing decisions, and secret handling on the server.
Write the audio direction
Describe the spoken content, voice style, background, music, scene timing, and any reference markers in the prompt field.
Choose reference mode
Use text only, add audio_urls for short reference clips, or send one image_url for visual guidance.
Set output controls
Pick output_format, sample_rate, speed, volume, and pitch so the request matches your downstream workflow.
Enable live generation safely
After sign-in, server configuration, credits, entitlement, and rate limits are approved, the API route can call https://fal.run/bytedance/seed-audio-1.0 and return the audio file object.
What Seed Audio 1.0 Can Help Create
Seed Audio 1.0 is useful when a prompt needs to describe more than a single voice line.
Launch videos
Generate narration, pacing, and a compact audio bed for product teasers and explainer clips.
Dialogue prototypes
Draft multi-speaker scenes for scripts, lessons, demos, and storyboards before recording final takes.
Audio brand studies
Explore how a product should sound with reference clips, voice direction, ambience, and music style.
Game and character tests
Prototype voices, creature moments, environmental cues, and short interactive audio states.
Podcast and course drafts
Turn outlines into listenable sections that help teams review rhythm and clarity early.
Developer API evaluation
Inspect the request fields, model ID, output file schema, and server-side key handling before integration.
Compare Seed Audio 1.0 with Miso One Voice Tools
Seed Audio 1.0 is one model in a larger AI voice workflow. Use these pages to compare text to speech, cloning, voices, pricing, and model analysis.
Seed Audio 1.0 FAQ
Practical answers about the Seed Audio 1.0 API, inputs, references, and Miso One integration.
What is Seed Audio 1.0?
Seed Audio 1.0 is a ByteDance audio generation model available on FAL. It generates natural audio from a prompt and can optionally use reference audio URLs or one image URL as guidance.
What FAL model ID does this page use?
The API contract targets bytedance/seed-audio-1.0 and validates requests for https://fal.run/bytedance/seed-audio-1.0 from the server. Live provider calls require approved workspace controls.
Is Seed Audio 1.0 the same as text to speech?
It can synthesize spoken content, but the prompt can also describe wider audio direction such as ambience, music feel, speaker roles, and reference-guided delivery.
Can I use reference audio with Seed Audio 1.0?
Yes. The FAL schema supports up to three audio_urls, each referenced in the prompt by order with @Audio1, @Audio2, and @Audio3.
Can I send both image_url and audio_urls?
No. The FAL schema treats the image reference and audio references as separate modes, so this page validates that they are not combined.
Which output formats are supported?
Seed Audio 1.0 supports wav, mp3, pcm, and ogg_opus. The default UI selection is mp3 because it is easy to preview and share.
Do I need to sign in to generate from this page?
Yes. Generation must be routed through a signed-in Miso One workspace with server-side FAL configuration, credits, entitlement, and rate limits. The page does not expose FAL keys in the browser.
How much does Seed Audio 1.0 cost on FAL?
The FAL model page lists Seed Audio 1.0 at $0.075 per generated minute. Miso One keeps live provider calls behind workspace controls so usage and costs can be managed before enablement.
What does the API return?
The output contains an audio file object with url, content_type, file_name, file_size, duration, channels, sample_rate, and bitrate when those values are returned by FAL.
Try the Seed Audio 1.0 request flow
Start in the generator above, inspect the payload, and connect approved workspace controls before live audio generation.