Seed Audio 1.0 · all-in-one audio

Signal Lost

A failing starship. A 4% chance. Three voices, the klaxons, the score — every second of the sound you're about to hear came from a single text prompt, generated in one pass.

Core cap

seed-audio

Voices

3, one prompt

In the mix

VO · SFX · score

Passes

Cost

~$2 e2e

▶ The motion-comic — 6 krea-2-os panels cut to the Seed Audio track. Unmute for the full mix.

Listen this is the raw output of ONE seed-audio call

◉ single pass · voices + sfx + score · already mixed

signal-lost.mp3 — generated by seed-audio (bytedance/seed-audio-1.0) on sdk.daydream.monster. No per-line TTS. No separate music. No foley. No mixing pass.

One prompt in a whole produced scene out

This is the entire input. Speaker, emotional direction in parentheses, the lines, then the ambience and the score — one text block. Seed Audio cast three distinct voices, performed the emotional arc, placed the sound effects, scored it, and mixed it.

COMMANDER VESPER (calm, authoritative woman, 40s): "Pilot — how long until the reactor breaches?" PILOT KAI (young man, panicked): "Eighty seconds! The coolant lines are gone!" ARIA, the ship's AI (synthetic, eerily calm): "Survival probability: four percent." COMMANDER VESPER (steady, resolute): "Then we make four percent enough. Reroute power to the dampeners — now." Ambience: blaring emergency klaxons and electrical sparks. Score: tense orchestral strings rising to a hopeful swell.

▼ ONE PASS ▼

The cast three voices, defined by description alone

No voice actors, no reference recordings — each voice was described in words and Seed Audio built it. (It can also clone a voice from up to 3 reference clips, or derive one from a character image.)

Cmdr. Vesper

"Calm, authoritative woman, mid-40s; steady under pressure." — the anchor.

Pilot Kai

"Young man, panicked, breathing hard; finds resolve." — the arc.

ARIA · ship AI

"Synthetic, eerily calm, genderless; a flicker of humanity." — the contrast.

Illustrated krea-2-os · one locked style, six beats

What one pass replaces for a single 45-second scene

The old way	Seed Audio 1.0
Cast + direct 3 voice actors, book a booth	3 voices from one prompt, each with tone direction
Hire a foley artist for klaxons / sparks / room tone	SFX described in the same prompt, placed in the mix
Commission a composer for the underscore	Original score that builds + resolves, same pass
Pay a mixing engineer to balance it	Already mixed — one master track out
Days of work + a studio budget	One pass · ~one minute · pennies

The recipe repeatable — swap the scene, same pipeline

01 · SCENE

One prompt → one mix

Script + ambience + score in one block → seed-audio → a single mixed audio_url. The whole point.

02 · CAST + PANELS

Illustrate it

One locked style → krea-2-os portraits + beat panels (~3s each).

03 · CUT

Motion-comic

Ken-Burns the panels to the audio with ffmpeg → a watchable reel. Optional: lip-sync a close-up with sync-lipsync-v3.

Why it matters. Audio was the last part of a scene you couldn't fake — voices, performance, sound design and score each needed a person. Seed Audio collapses all four into one prompt, one pass, already mixed. A game studio scripts a cutscene; a novelist hears their chapter; a teacher builds a two-character dialogue — in the time it takes to write it.

Make your own. Open the One Prompt, a Whole Scene playbook, fill in a SCENE, and run it.