Give every character in your story a voice of their own — picked from a library, designed from a description, or cloned with permission — then direct each line the way you’d brief an actor.
Here’s how a short audio-drama scene comes together on Breeze Blue: each character gets an identity from one of three sources, every line gets its own direction, and the scene generates as clips you arrange and export.
Sketch each character in a line or two — age, temperament, how they carry themselves. This sheet becomes your voice choices and line directions.
Pick an identity from the Voice Library, design one from your character description, or clone an authorized voice. Save each as a named voice asset.
Attach a plain-language instruction to each line — pace, mood, subtext. The voice keeps its identity while the performance follows your direction.
Generate each line with its character’s voice and direction, arrange the clips in order in Studio, and export the finished scene as audio.
Sea captain, late 50s — weathered, unshakable, speaks in short orders.
Picked from the Voice Library: a low, grainy voice with a slow, deliberate cadence.
“Tie it off. We’re not losing another lantern to this storm.”
Low and steady, almost a growl — she has done this a hundred times.
Deckhand, 16 — bright, jumpy, talks faster when scared.
Designed from one line of description: “a reedy teenage voice that speeds up and cracks when nervous.”
“Captain — the rope! It’s already fraying!”
Panicked and breathless, pitch rising at the end of the line.
The storyteller — wry, unhurried, one step outside the action.
Cloned from the creator’s own authorized voice sample, so the show keeps its host’s sound.
“The storm, as it turned out, had plans of its own.”
Calm and wry, with a small pause before “plans.”
Each line generates as its own clip — Mara’s growl, Pip’s panic, the narrator’s dry aside — carrying its character’s saved voice and its line-level direction. In Studio you arrange the clips in order, regenerate any read you want tighter, and export the scene as one audio file.
Every cast is a mix. The right source depends on whether the voice already exists, lives only in your head, or belongs to someone who has given you permission to use it.
Fastest when a curated voice already fits. Browse voices with real character — each one fully directable — audition your lines on them, and save the winner to your cast.
Best for: supporting roles, narrators, and getting a scene drafted today.
Explore the public Voice Library →When the character exists only on the page, describe them — “a reedy teenage deckhand who cracks when nervous” — and hear an original voice made to match. No reference audio needed.
Best for: original protagonists, NPCs, creatures, and voices that shouldn’t sound like anyone.
See how AI voice design works →When the voice belongs to you or to someone who has consented — a host, an actor you work with — clone it from a short authorized sample (5 seconds of clean audio is enough; we analyze the first 60 seconds and keep the best 30) and cast it like any other voice.
Best for: your own narrator voice, recurring hosts, and actors on your team.
Read the voice cloning workflow →The mistake most casts suffer from is conflating who a character sounds like with how they deliver a given line. Breeze Blue keeps the two separate on purpose.
A character’s timbre — the grain, register, and texture that make them recognizable — comes from the voice you picked, designed, or cloned. You choose it once, save it, and it stays stable from the first line of episode one to the last line of the season.
Everything else — pace, mood, subtext, restraint — is direction, written per line in plain language on top of the same voice. Mara can growl an order, mutter a doubt, and break down in grief without ever becoming a different person.
This separation is measured, not assumed: in our Voice Direction benchmark (August 2026), Breeze TTS 2 scores 4.25 overall on direction-following while holding speaker similarity at 0.67 — the performance moves, the identity doesn’t.
Voices you save on Breeze Blue belong to your account, not to a single project. Build the cast once and bring it anywhere.
Library picks, designed voices, and authorized clones all save to your collection as named assets — Mara stays Mara across scripts, episodes, and rewrites.
The same saved voice is available in single-line Text to Speech and in Studio’s multi-voice scripts, so a quick pickup line and a full scene draw on the identical identity.
Each saved voice has an ID you can reference from the Developer API, so the cast you assembled in the browser performs in your pipeline, game build, or batch render.
Reuse is deliberate: you choose which saved voice each project and each line uses. Breeze Blue doesn’t recast or reorganize your collection behind your back.
Write or paste a scene, assign each line a character voice from your saved cast, direct the reads, and export finished audio — the workflow behind the walkthrough above.
Every saved voice is addressable by ID from the Developer API: send a line, a voice, and a plain-language direction, and stream back the performed audio for games, agents, and batch pipelines.
A character voice generator creates a distinct synthetic voice for a fictional character and performs their lines in it. On Breeze Blue that means two decisions: an identity — picked from the Voice Library, designed from a text description, or cloned from an authorized sample — and a per-line performance you direct in plain language, the way you’d brief a voice actor.
Yes. Voice Design builds an original voice from a written description alone — age, texture, attitude, accent — so characters that exist only in your script or concept art can get a voice without any source recording. Library picks also need no audio from you.
Clone when the voice already exists and you have permission to use it — your own narration, a recurring host, an actor you work with. Cloning needs a short authorized sample (5 seconds of clean audio is enough; we analyze the first 60 seconds and keep the best 30) and a consent confirmation. For characters who exist only on the page, designing a voice or picking one from the library is the better fit.
Save the voice as a named asset and reuse that same asset everywhere the character speaks. The identity stays fixed while your line-by-line direction changes the performance — so episode ten’s Mara is the same Mara, just angrier, softer, or more tired as the scene demands.
Direction changes the performance, not the person — a single voice can whisper, shout, and hesitate, but it keeps its recognizable timbre. For genuinely different characters, give each one its own voice; it costs a description or a library pick, and your scenes stop depending on the listener not noticing.
Yes. Voices you save in the browser are addressable by ID from the Developer API: send the text, the voice, and a plain-language instruction, and get the performed audio back. That covers game dialogue pipelines, agents that speak in character, and batch narration jobs.
Audio generated on paid plans can be used commercially — in games, videos, audiobooks, and podcasts. Breeze Blue is free to start with free credits you can claim daily, and paid plans add monthly credits and voice slots; the full breakdown is on the pricing page.
