01 / 15sCINEMATIC DIALOGUE
Suspense dialogue package
A rain-soaked crime scene shaped with intimate narration, environmental tension, and a precisely timed final cue.
Seed Audio 2.0 model notes, playable examples, and prompt limits →
✦ Model introduction
SEED AUDIO 2.0How Seed Audio 2.0 builds an audio scene
Give it a scene brief, not just a line to read. Describe the speakers, pacing, room, music, and key sound events; the model returns them as one draft.
Text, image, audioPrompt and reference inputs
Layered scenesVoice, music, ambience, and foley
Up to 2 minutesOne continuous output window
AUDIO COLLECTION
Listen before you write a long prompt
These short clips show what changes when a prompt specifies speakers, pacing, setting, and a few important sound cues.
SEED AUDIO 2.0 LISTENING ROOM
Six short examples, with the prompts kept practical.
Each clip focuses on one job: dialogue, livestream delivery, podcast pacing, an ad read, voice reference, or layered scene sound.
01 / 15s
02 / 14sLIVESTREAM VOICE
Commerce livestream voice
Two energetic hosts move from product detail to a confident call to action with clean, broadcast-ready pacing.
03 / 15sCONVERSATION
Conversational podcast
A warm, natural podcast opening with thoughtful pauses, a grounded delivery, and an intimate studio presence.
04 / 14sMULTI-ROLE SPOT
Multi-role ad spot
A compact commercial concept that moves through narrator, customer, and brand voice in one cohesive sequence.
05 / 15sVOICE REFERENCE
Controlled voice reference
A calm reference-led performance that keeps tone, rhythm, and spoken texture consistent across the generated scene.
06 / 16sLAYERED OUTPUT
Multi-reference scene
Dialogue, weather, score, and foley are directed as coordinated layers for a richer Seed Audio 2.0 production draft.
SHORT ANSWER
TTS reads a script. This builds the room around it.
Traditional text-to-speech focuses on spoken words. Seed Audio 2.0 expands the creative direction to the entire scene: character delivery, timing, space, background music, ambience, and sound effects can be composed as one connected experience.
CAPABILITY MAP
What Seed Audio 2.0 is useful for
It works best when speech and setting need to arrive together. For plain narration, a standard TTS tool may be simpler.
Multimodal direction
Begin with a text prompt, image reference, or short audio reference to guide the scene’s mood, rhythm, and sonic identity.
Dialogue and delivery control
Describe speakers, tone, pacing, emotion, accents, pauses, and conversational energy directly in the prompt.
Layered sound design
Plan dialogue, ambience, music beds, transitions, and precisely timed sound events inside one coherent direction.
Consistent creative workflows
Reuse references and prompt structures to keep recurring characters, campaigns, and series aligned across scenes.
MODEL NOTES
Useful constraints before designing a prompt
Keep the scene concise, specific, and ordered around audible events rather than visual-only details.
Use one clear image or a small set of focused audio references so each source has an obvious role.
Trim references to the voice, texture, rhythm, or performance quality you actually want to preserve.
Draft shorter scenes first, then extend once dialogue timing and layer balance feel right.
State the spoken language, desired accent, pronunciation notes, and any speaker changes in the prompt.
Choose the delivery format and sample rate that match editing, publishing, or product integration needs.
Review speed, pitch, volume, and voice choices before generation to avoid unnecessary revisions.
Treat generation as an asynchronous production step: submit, track progress, validate, then deliver.
WHERE IT FITS
Good fits—and a reason to prototype first
Seed Audio 2.0 is designed for ideas that need voices, atmosphere, music, and sonic action to arrive as one connected scene.
Short video and ads
Prototype campaign hooks, localized voiceovers, social clips, and complete sound beds before final production.
Audiobooks and audio drama
Shape narration, character delivery, room tone, music, and lightweight foley from one scene direction.
Games and interactive media
Explore character voices, ambience loops, interface moments, and cutscene audio during early production.
Podcasts and explainers
Turn a topic into a guided audio sketch with natural pacing, transition cues, and background texture.
PROMPT FRAMING
A practical Seed Audio 2.0 prompt structure
Build the direction in four passes so the model understands the story first, then the performance and sound design.
Scene
Name the format, place, mood, duration, and intended audience.
Voice
Define speaker count, language, role, emotion, pacing, and pronunciation.
Layers
Add ambience, music direction, and only the sound events that matter.
Review
Generate a short draft, identify the dominant layer, and revise one variable at a time.
ACCESS NOTES
Try it, then check the details
Try the Seed Audio 2.0 generator↗
Turn a focused scene direction into a playable audio draft in the web workspace.
Read the prompt guide↗Use tested structures, practical examples, and revision checks before writing longer prompts.
Explore API documentation↗Review request fields, output options, and integration notes for production workflows.