Fish Voice Designer
Describe a voice in plain words and get it as a reusable AI voice, ready to narrate anything.
Fish Voice Design
Describe a voice in plain words and get it as a reusable AI voice, ready to narrate anything, in your browser on mitte.ai.
Fish Voice Design invents a voice from a description. No recording, no voice actor, no sample to upload. You write what the voice should sound like, optionally add a thumbnail, and a few seconds later the voice sits under My presets with a sample clip, ready to read any script you give it. It runs on Fish Audio’s voice design model.
What is voice design?
Voice design generates a brand-new synthetic voice from a natural-language brief. You describe the speaker: age, gender, tone, accent, pacing, attitude. The model composes a voice that matches and reads a short preview line so you can hear it immediately.
It differs from voice cloning in where the voice comes from. Cloning copies a real person’s voice from a recording; design invents someone who doesn’t exist. That makes it the right tool when you need a character, a brand voice with no face attached, or narration without licensing anyone’s actual voice.
On mitte the designed voice becomes a preset. Open it, type your script, and it speaks, with the same emotion tags, speed control and 13 languages as any Fish Text-to-speech run.
Key features
- A sentence is enough: “Warm, confident studio narrator with a natural tone” already produces a usable voice. More detail sharpens the result.
- Ready in seconds: no casting, no session, no training queue.
- Saved as a preset: the voice lands under My presets with a name, description and sample clip filled in for you, private to your account.
- Speaks 13 languages: the designed voice keeps its character in English, Spanish, Japanese, Arabic and 9 more.
-
Full delivery control: direct the performance with emotion tags like
[happy],[whispering]and[sarcastic]when you generate speech with it. - No rights questions: the voice belongs to nobody, so there’s no one to clear it with.
How to design a voice on mitte.ai
- Open Voice Design on mitte.ai.
- Describe the voice. Age, gender, tone, accent, pacing, context. Example: “Calm female documentary narrator, low register, unhurried, slight British accent.”
- Add a thumbnail if you want a cover image for the voice. This step is optional.
- Run it. The voice appears under My presets a few seconds later, with a sample clip so you can hear it right away.
- Use it. Open the preset, paste your script, pick a language and generate speech.
How to control emotions?
The brief sets the voice’s baseline character; the delivery of each line is controlled later, when you generate speech with the preset. Place cues in square brackets where you want the performance to change:
[excited] Introducing our newest product!
[calm] Take a deep breath and begin.
[angry][shouting] Get out of here now!
A tag at the start of a sentence sets its emotion; you can layer tags for combined effects, add human sounds like [laughing] and [sighing], or write free-form descriptions like [warm and happy]. There are 64+ expressions in total. See the full tag reference on the Fish Text-to-speech page.
Writing a good voice brief
The description is the whole input, so it earns some care:
- Name the delivery, not just the person. “Fast-paced, high energy, persuasive” gives the model more than “young man.”
- Anchor it in a context. “Late-night radio host”, “kids’ show narrator”, “luxury brand commercial” each pull the voice in a specific, useful direction.
- Pin the accent and pace if they matter: “slight Irish accent, measured pace” beats hoping for the best.
- One voice per brief. Describe a single speaker; contradictory traits (“booming whisper”) muddy the result.
A brief that works: “Excited male product reviewer, mid-30s, fast-paced, high energy, and persuasive.”
What creatives use it for
- Character voices. Cast an entire animation or game from text: the gruff mentor, the nervous sidekick, the villain who whispers.
- Brand voices. Design a voice that fits the identity, use it across product videos, ads and tutorials, and never worry about a voice actor’s availability.
- Narration without casting. Documentary, explainer or audiobook narration matched to the material’s mood in one brief.
- Scratch VO with intent. Temp narration that already sounds like the final direction, so the edit sells before the recording session exists.
- Localization-ready voices. One designed voice reads the campaign in 13 languages with the same character.
FAQ
What do I need to provide? Just a written description of the voice, up to 2,000 characters. A thumbnail image is optional. No audio sample is involved.
How long does it take? Seconds. The preset appears under My presets with a sample clip as soon as the run finishes.
How is this different from voice cloning? Cloning reproduces a real voice from a recording and needs a sample. Design invents a new voice from words alone. If you want your voice, use the Voice Cloner; if you want a voice, design it.
Can the designed voice speak other languages? Yes. Like every Fish voice on mitte, it speaks 13 languages: English, Chinese, Japanese, German, French, Spanish, Korean, Arabic, Russian, Dutch, Italian, Polish and Portuguese.
Can I control the emotion and delivery?
Yes. Generate speech with the preset and use the full set of inline tags: [happy], [angry], [whispering], [sighing] and 60+ more, documented on the Fish Text-to-speech page.
Where does the voice live? Under My presets, private to your account. Reuse it in any project by opening the preset and typing new text.
Can I use it commercially? The voice is synthetic and belongs to no real person, which removes the consent questions cloning carries. Check your project’s usual licensing requirements as you would for any generated media.
Got a character in mind? Design their voice on mitte.ai.