Guide

Audio Prompting Guide — Write Prompts Like a Screenwriter

Seed Audio 1.0 turns one prompt into dialogue, music, ambience and effects. The prompts that work read like tiny screenplays. This guide shows the structure, the reference syntax, and eight copy-ready examples.

Open the workbench

The script-style structure: who, feeling, scene, sound

Every reliable scene prompt answers four questions. WHO is speaking — name each character and give them a voice ("male, gravelly, tired"). WHAT they feel — emotion drives the performance more than any adjective about audio quality. WHERE it happens — the space defines reverb, distance and background ("a stone cavern" vs "a car interior"). WHAT ELSE sounds — call out music, ambience and effects explicitly, or the model has to guess.

Write dialogue lines in quotes, in the language you want spoken. Keep each generation to one coherent beat of a story — two to four characters, one location — and let Extend carry the plot forward.

@Audio1–@Audio3: voice cloning by reference

Upload up to three reference clips alongside your prompt, then assign them by writing @Audio1, @Audio2 or @Audio3 wherever a voice is cast: "The detective speaks with @Audio1's voice." Each reference clones the timbre of one voice, so three clips can cast three distinct characters in a single scene.

Clean, dry recordings clone best: a quiet room, no music, a few natural sentences. The clone follows your prompt's emotional direction — a calm reference can still deliver an angry line.

Character images — the other way to cast (mutually exclusive)

Instead of reference audio, you can attach one character image. The model reads age, build, attitude and setting from the picture and invents a fitting voice — great when you have concept art but no voice actor. The two casting methods are mutually exclusive: a prompt uses @Audio references or one image, never both.

Directing TTS with voice_instruction

Text to speech (Seed Speech 2.0) separates the script from the direction. The script is read verbatim; voice_instruction is a private note to the voice — tone, emotion, pacing, attitude — and is never spoken. "Read like a flight attendant announcement", "barely holding back laughter", "grave documentary gravitas": same script, radically different takes.

Combine it with the mechanical controls — speed, pitch, volume — for the final polish. Directing lives in the instruction; physics lives in the sliders.

Extend: the strategy for anything longer than two minutes

Scene generations run up to two minutes. Extend continues an existing piece with the same character voices, music bed and mix — so plan long work as a chain: write beat one, generate, audition, then Extend with a prompt that only describes what happens next. Do not re-describe the cast or the room; the continuation inherits them.

Iterate cheaply at the head of the chain: fixing the opening beat before extending saves regenerating everything downstream. For audiobooks, extend chapter by chapter; for ambience, extend into long seamless loops.

Eight copy-ready example prompts

1. Two-character radio drama scene

EN
A 1940s detective office at night. MARLOW (male, gravelly, tired) interrogates VERA (female, young, defensive but hiding fear). Rain taps the window, a desk clock ticks, low noir strings underneath. Marlow: "Where were you at midnight?" Vera hesitates, then answers with a tremor.
中文
深夜的 1940 年代侦探事务所。马洛(男声,沙哑疲惫)审问薇拉(女声,年轻,嘴硬但藏着恐惧)。雨点敲着窗户,桌上时钟滴答,底下铺低沉的黑色电影弦乐。马洛:"午夜十二点你在哪里?"薇拉迟疑片刻,带着颤音回答。

2. Podcast intro in your own cloned voice

EN
Podcast intro, 15 seconds. The host speaks with @Audio1's voice: "Welcome back to Night Shift — stories from the city after dark." Warm and unhurried, over a mellow lo-fi beat with soft vinyl crackle that fades under the last word.
中文
播客片头,15 秒。主持人用 @Audio1 的声音说:"欢迎回到《夜班》——打烊之后的城市故事。"语气温暖从容,底下是柔和的 lo-fi 节拍加轻微黑胶底噪,在最后一个字时渐弱。

3. Full cast from three reference voices

EN
Family kitchen, morning chaos. DAD speaks with @Audio1's voice (calm, deadpan), MOM with @Audio2's voice (rushed, warm), SON with @Audio3's voice (sleepy teenager). Sizzling pan, radio jingle in the background, chair scraping. They argue about who finished the coffee.
中文
清晨鸡飞狗跳的家庭厨房。爸爸用 @Audio1 的声音(平静冷面),妈妈用 @Audio2 的声音(匆忙温暖),儿子用 @Audio3 的声音(睡不醒的高中生)。平底锅滋滋作响,背景收音机放着广告歌,椅子擦地。三人为谁喝完了最后一杯咖啡拌嘴。

4. Anchoring a character with an image

EN
[Attach one character image — no @Audio references in the same prompt.] The knight in the image narrates her own legend around a campfire: weathered female voice, quiet pride, crackling fire, night crickets, a distant wolf howl at the end.
中文
[附一张角色图片——同一条 prompt 中不能再用 @Audio 引用。]图中的女骑士在篝火旁讲述自己的传说:饱经风霜的女声,压着一点骄傲,柴火噼啪,夜虫鸣叫,结尾远处一声狼嚎。

5. TTS ad read, directed with voice_instruction

EN
Script: "Meet Bloom — the planner that plans around your life. Try it free this week." voice_instruction: energetic launch-day announcer, smiling, punch the word "free", quick confident close.
中文
正文:"认识一下 Bloom——围着你的生活做计划的日程工具。本周免费试用。" voice_instruction:发布日主持人的亢奋感,带笑,重读"免费",收尾干脆自信。

6. TTS bedtime story at reduced speed

EN
Script: "The little fox curled up under the pine tree, and the snow kept falling, soft as sleep." voice_instruction: intimate bedtime storyteller, almost a whisper, long gentle pauses. Speed: 0.85x.
中文
正文:"小狐狸蜷在松树下,雪一直落着,轻得像睡意。" voice_instruction:贴耳的睡前讲述者,近乎耳语,停顿绵长温柔。语速:0.85 倍。

7. Loopable game ambience

EN
Loopable dungeon ambience, 90 seconds: slow water drips echoing in a stone cavern, faint wind through cracks, occasional distant chain rattle. No music, no voices, no one-off events — must loop seamlessly.
中文
可循环地牢环境音,90 秒:石窟中回响的缓慢滴水,风从石缝挤过的细响,偶尔远处铁链轻晃。不要音乐、不要人声、不要一次性事件——必须无缝循环。

8. Extend: continuing a scene consistently

EN
[Use Extend on the previous generation, then prompt only what happens next.] Vera finally confesses. Same voices, same rain and clock. The strings swell slightly as she says: "I was there. But so was someone else."
中文
[对上一段生成使用 Extend,只描述接下来发生的事。]薇拉终于开口坦白。声音、雨声、时钟都保持原样。弦乐微微上扬,她说:"我在场。但在场的不止我一个。"

Common mistakes checklist

  • Vague sound words. "A door sound" could be forty different doors — name the material, the space, and the energy: "heavy iron gate creaking open in a stone corridor".
  • Putting stage direction inside a TTS script. In text to speech, everything in the script gets read aloud — direction belongs in voice_instruction, which is never spoken.
  • Mixing @Audio references with a character image. They are mutually exclusive — pick voice cloning (up to 3 clips) or one image per prompt, not both.
  • Overcrowding a scene. Ten characters in a two-minute generation blur together; two to four distinct voices per scene stays clean.
  • Regenerating instead of Extending. A fresh generation recasts the voices; Extend continues the exact same voices, music and mix.
  • Forgetting "loopable" for ambience — and describing one-off events (a scream, a crash) inside something meant to repeat.
  • Changing voice or instruction mid-project. For audiobooks and series, save the exact voice + voice_instruction pair and reuse it verbatim every chapter.

Prompting FAQ

Put it into practice

The best way to learn prompting is to generate

Copy an example above, paste it into the workbench, and iterate. Free credits on signup.

Try it free