Character voice
You can design a pirate voice with one prompt
Describe the character, the age range and the delivery in one sentence. Gradium returns a voice id you call from the API or the SDK, with no actor, no session fee and no licensing. The prompt below is the one that produced the voice in the sample.
Hear it
- PirateEnglish · Bristol0:00
“Listen here, ye sea-bitten scallywags. We've sailed through the devil's own squalls and danced with the Kraken itself, but the prize that lies beyond that reef is worth every drop of sweat and blood.”
The prompt
that produced it
A gruff Bristolian English male pirate voice, 45 to 60, for game and character narration: weathered low pitch, gravelly timbre with heavy vocal fry, strong projection, at a steady, unhurried pace, boisterous and commanding energy.
This exact wording produced the voice in the sample above. Change any part of it and the voice changes with it.
- Pitch1/5
- lowhigh
- Pace2/5
- slowfast
- Energy5/5
- calmboisterous
This prompt
in four styles
A prompt is built from parts: the age, then pitch, texture, projection, pace and energy. Change one part and you get a variation. These are starting points to adapt, not finished prompts.
The prompt as written. Low pitch, unhurried pace, boisterous energy, and the vocal fry carrying the weathering.
A Bristolian English male voice, 45 to 60, for game and character narration: gruff, with weathered low pitch, steady, unhurried pace, boisterous and commanding energy, gravelly timbre with heavy vocal fry and strong projection.
Copy this as it stands.
Take the energy out and keep everything else. Menace is quiet, which is the opposite of what most character prompts ask for, and it is why villain voices so often come back sounding like pantomime.
A Bristolian English male voice, 45 to 60, for game and character narration: cold and deliberate, with weathered low pitch, steady, unhurried pace, low energy, gravelly timbre with heavy vocal fry and a quiet, close delivery. Ideal for antagonists, boss dialogue and horror narration.
In colour: the 4 parts that make this style.
Push the age up and swap gruff for warm. Watch intelligibility here: older voices drift breathy, so add clear diction if the character has to deliver instructions.
A Bristolian English male voice, 60 to 75, for game and character narration: warm and worn, with weathered low pitch, steady, unhurried pace, low-to-mid energy, gravelly timbre with heavy vocal fry and strong projection. Ideal for quest-givers, narrators and tutorial guides.
In colour: the 4 parts that make this style.
The one direction where fast pacing works, because the character is meant to trip over itself. Keep the vocal fry or it stops sounding like it belongs in the same world as the others.
A Bristolian English male voice, 25 to 40, for game and character narration: eager and scattered, with mid-high pitch, fast pacing, boisterous and commanding energy, gravelly timbre with heavy vocal fry and strong projection.
In colour: the 4 parts that make this style.
What this prompt
is actually doing
Five things worth copying into any character voice, and the reasons they are in this prompt rather than left to chance.
- Name the region, not the trope
- This prompt asks for Bristolian English, not for a pirate accent. The stage-pirate voice is really a West Country one, and naming an actual place gives the model something acoustic to aim at. Asking for a pirate accent instead gets you an impression of an impression, which falls apart the moment the script says anything ordinary.
- Vocal fry is the texture doing the work
- Gravelly timbre with heavy vocal fry is what reads as weathered. It is a texture instruction rather than an emotion, which matters because textures hold across a long script while emotions drift. If a character voice sounds right for one line and wrong for the next, the prompt is usually carrying emotion where it should carry texture.
- Projection is not volume
- Strong projection asks for the posture of someone filling a room, not for a louder file. Loudness is a mastering decision made after generation; projection changes how the voice is produced, which is audible in the resonance even at low volume.
- Energy 5 needs pace 2 to stay intelligible
- The pitch/pace/energy triplet here is 1/2/5, and the low pace is not incidental. Boisterous energy at a fast pace turns into shouting and loses consonants; keeping the pace unhurried is what lets the energy read as commanding rather than frantic. This is the most useful thing to copy from this prompt into any other character voice.
- Character voices need an audition script with range
- The showcase script is deliberately theatrical, but a character voice you intend to ship should also be auditioned on something flat and administrative. A pirate voice reading a menu is how you find out whether you have a voice or a performance.
The prompt,
and where it goes wrong
The method behind the prompt: what each clause does, what breaks it, and what it was tested on.
Each row is one part of the prompt above and what it is set to here. Change a single row and you get a variation on this voice; change several and you get a different one.
| Clause | This prompt |
|---|---|
| Accent / locale | Bristolian English |
| Gender | male |
| Age range | 45 to 60 |
| Purpose | game and character narration |
| Character | gruff |
| Pitch | weathered low pitch |
| Pace | steady, unhurried pace |
| Energy | boisterous and commanding energy |
| Resonance / timbre | gravelly timbre with heavy vocal fry |
| Modifiers | strong projection |
4 failure modes we hit repeatedly, and what fixes each one.
- Asking for a pirate accent rather than a real regional one produces a caricature that collapses on ordinary sentences.
- Name Bristolian or West Country English and let the character come from the energy and texture clauses.
- Boisterous energy combined with fast pacing turns into shouting, and consonants disappear at the ends of words.
- Hold the pace at a steady, unhurried value. The 1/2/5 triplet works precisely because the pace stays low while the energy is high.
- Stacking several character adjectives at once produces a different voice on each generation, because the model has too many targets to satisfy.
- Change one part at a time. Gruff and weathered and boisterous and gravelly is four instructions competing for the same job, where the prompt above gives pitch, texture, projection, pace and energy each their own words.
- A character voice that sounds perfect on the demo line falls apart across a game script, because the prompt encoded a performance rather than a voice.
- Audition every candidate on a deliberately flat line as well as the dramatic one. Anything that only works on the dramatic line is a read, not a voice.
Listen here, ye sea-bitten scallywags. We've sailed through the devil's own squalls and danced with the Kraken itself, but the prize that lies beyond that reef is worth every drop of sweat and blood.
Deliberately theatrical, because that is what a character voice has to survive. Audition every candidate on something flat and administrative as well, or you will not find out whether you have a voice or a performance until the script is written.
Descriptor words do not transfer between locales. Energetic and bright travel intact, but most descriptors resolve to a different sound depending on the language they are read in. Character work is the most exposed to this, because a trope like the stage pirate is culturally specific: the French equivalent is a different set of acoustic cues, not a translation of these words.
Questions?
Answered
"A gruff Bristolian English male pirate voice, 45 to 60, for game and character narration: weathered low pitch, gravelly timbre with heavy vocal fry, strong projection, at a steady, unhurried pace, boisterous and commanding energy." That is the exact prompt that produced the voice in the sample above.
Because the stage-pirate accent is really a West Country one, and naming a real place gives the model an acoustic target instead of a trope. Bristolian is not in our mapped accent set, which is why there is no Bristolian accent page: the model approximates it here in service of a character, and that is a different promise from offering it as an accent.
Yes. A designed voice is generated rather than cloned from a person, so there is no actor to license and no session to re-book when the script changes. That is the practical reason studios use Voice Design for NPC dialogue: hundreds of distinct voices without hundreds of casting calls.
Keep the character in the texture clauses rather than the emotional ones. Textures such as gravelly timbre and vocal fry hold across arbitrary new text; emotional instructions drift line to line. Then audition the candidate on a flat, ordinary sentence before you commit to it.
Design your own
character voice
A designed voice is a normal voice id: same endpoints, same SDK, same latency as any library voice. Every voice we ship clears blind listener comparison and an ELO ranking first, because roughly one candidate in a thousand has it.
Paste the prompt, change the parts that matter to you, and generate. No casting call, no session fee, no licensing.