AI Voice Design: Create a Custom AI Voice from a Description
VdoBloom AI Voice Design creates a brand-new synthetic voice from a written description. Type something like "a warm, deep male narrator with a slight British accent, calm pace", hear it read a sample line, and the voice is saved to your account. Then switch to Speak and use that same voice to read any script up to 2,000 characters, at any speed from 0.5× to 2×.
Is it free? New accounts get 10 free credits, and voice design is priced by the characters spoken: a design run with its sample line costs 1 credit, and reading a full 2,000-character script costs 3. So the free credits are enough to design several voices and voice a few scripts. After that, 300 credits are $19.50 as a one-time pack that never expires, or $15 a month on a plan.
It is made for YouTubers and faceless channels that want a consistent narrator, game and story creators who need distinct character voices, marketers voicing ads, and anyone who cannot find the right voice in a stock list. Voices are generated from your words only; uploading or cloning a real person's voice is not supported.
How it works
- 1
Describe the voice
Write who is speaking: age, gender, accent, tone, pitch, pace and personality, in 5 to 1,000 characters.
- 2
Set a sample line, name and language
Keep the default sample sentence or write your own (up to 300 characters), give the voice a name such as "Narrator", and pick a language or leave it on Auto.
- 3
Design the voice and listen
Press Design Voice. The sample plays in the results panel, downloads as an MP3 and the voice is added to your list. Not quite right? Adjust the description and design again.
- 4
Speak with it
Open "2. Speak with it", pick your designed voice, paste up to 2,000 characters of script, set the speed and press Speak to get the full voiceover.
Why use this tool
Voice from words, not recordings
No microphone, sample or voice actor needed. The voice is built entirely from your text description.
Reusable saved voices
Every designed voice is stored on your account with its sample, so you can come back and voice new scripts in exactly the same voice.
10 languages plus Auto
Design and speak in English, Chinese, Japanese, Korean, German, French, Russian, Portuguese, Spanish or Italian, or let the model detect the language from the text.
Speed control
Read scripts anywhere from 0.5× to 2× speed, in 0.05 steps, to fit a video length or a character's energy.
Pay per character
Short lines cost 1 credit and a full 2,000-character script costs 3, so you only pay for what the voice actually says.
Safe by design
Only voices you designed in the tool can be used to speak. There is no upload path, so nobody's real voice can be copied through it.
How to describe a voice: attributes that matter
The model listens for concrete traits. Cover a few of these in one or two sentences and the result gets much closer to what you imagine:
| Attribute | What to write | Example words |
|---|---|---|
| Age | Roughly how old the speaker sounds | young, mid-30s, elderly |
| Gender and pitch | Voice register | deep male, bright female, low, husky |
| Accent | Region or dialect | slight British, Southern American, Australian |
| Tone | Emotional colour | warm, authoritative, playful, sinister |
| Pace | Speaking speed and rhythm | calm, slow and deliberate, fast and energetic |
| Texture | Physical quality of the voice | raspy, breathy, smooth, gravelly |
| Role | What the voice is for | documentary narrator, radio host, fantasy villain |
Example voice descriptions to start from
Copy one of these into the description box and change the details to suit your project:
- Documentary narrator: a warm, deep male voice in his fifties, slight British accent, calm and measured pace.
- Product explainer: a friendly, clear female voice in her late twenties, neutral American accent, upbeat but not rushed.
- Bedtime story: a soft, gentle grandmother with a slow, soothing pace and a smile in her voice.
- Game villain: a cold, gravelly male voice, low pitch, slow and menacing, with long pauses.
- Energetic host: a young, enthusiastic male presenter, fast pace, bright and punchy, perfect for short-form intros.
- Calm meditation guide: a breathy, relaxed female voice, very slow pace, quiet and reassuring.
Read more: Guide: designing AI character voices
How much does AI voice design cost?
Both designing and speaking are billed at 0.0012 credits per character, rounded up, with a 1-credit minimum. In practice:
| What you run | Characters | Credits |
|---|---|---|
| Design a voice (with the default or any sample line) | up to 300 | 1 |
| Speak a short script (about 130 words) | up to 833 | 1 |
| Speak a medium script (about 250 words) | 834–1,666 | 2 |
| Speak a full script (about 330 words) | 1,667–2,000 | 3 |
Read more: Credit packs and plans
Is AI voice design free?
You can start free. The 10 credits every new account receives cover up to ten voice designs, or a few designs plus several full-length scripts. There is no daily allowance after that, but one-time credit packs never expire, so a 300-credit pack for $19.50 covers 300 design runs or 100 full 2,000-character scripts.
Voice Design or Text to Speech?
Both turn text into speech. The difference is where the voice comes from.
| AI Voice Design | Text to Speech | |
|---|---|---|
| Voice source | A new voice you describe in words | A ready-made voice picked from a list |
| Best for | A unique narrator or character voice you reuse | Quick voiceovers with a stock voice |
| Reuse | Saved to your account and picked again under Speak | Pick the same stock voice each time |
Read more: Open Text to Speech
What creators use custom AI voices for
A designed voice is most useful when you need the same voice again and again.
- Faceless YouTube and TikTok channels with one recognisable narrator across every video.
- Audio dramas, games and animations where each character needs a distinct voice.
- Ads and explainers in a brand voice that stays consistent between campaigns.
- Audiobook and story samples in several languages from one description.
- Prototyping voice lines before hiring a voice actor.
Read more: How to add an AI voiceover to a video
Limits to know
Each Speak run reads up to 2,000 characters; split longer scripts into parts and join them in your editor. Sample lines are up to 300 characters. The tool designs voices from descriptions only: it does not clone, imitate or accept recordings of real people, including celebrities. Descriptions and scripts go through content moderation, and if a run fails its credits are refunded automatically.
Frequently asked questions
What is AI voice design?
AI voice design creates a new synthetic voice from a text description of how it should sound, instead of choosing from a stock list or cloning a recording. On VdoBloom you describe the voice, hear a sample, and then use it to read any script.
How do I create a custom AI voice?
Open AI Voice Design, describe the voice (age, accent, tone, pace), keep or change the sample line, pick a language and press Design Voice. Once you like it, go to "Speak with it" and paste your script.
Is AI voice design free?
New accounts get 10 free credits. Designing a voice costs 1 credit, and reading a script costs 1 to 3 credits depending on length, so you can try it without paying.
Can I clone my own voice or a celebrity's voice?
No. Voice Design builds voices from descriptions only. There is no way to upload a recording, so neither your voice nor anyone else's can be cloned with this tool.
What languages are supported?
English, Chinese, Japanese, Korean, German, French, Russian, Portuguese, Spanish and Italian, plus an Auto setting that detects the language from the text.
Can I reuse a voice I designed?
Yes. Every designed voice is saved to your account with its sample. Choose it under "Speak with it" whenever you need a new script read in the same voice.
How long can the script be?
Up to 2,000 characters per Speak run, which is roughly 330 words. For longer narration, generate it in parts and join the MP3 files.
Can I change how fast the voice talks?
Yes. The speed slider goes from 0.5× to 2×. You can also write the pace into the voice description, such as "slow and deliberate".
Why does the voice not sound like my description?
Short or abstract descriptions leave the model guessing. Name the age, accent, tone and pace, add a texture word such as raspy or breathy, and write a sample line that suits the character. Then design again; each run costs about 1 credit.
What format is the audio?
Samples and scripts download as MP3 files and are also saved to My Creations.