AI Text to Speech

1. 1

   Paste the script. Punctuation does real work here: a comma and a full stop are different pauses, and the delivery follows them.

2. 2

   Pick a voice from the catalogue, filtered by language, gender and accent, and audition the preview before you spend a credit.

3. 3

   Tune stability, similarity, style and speed, then generate. Stability trades consistency against expression — low is more performed, high is more even.

Three prompts that work well. Open the generator and paste one in.

• Welcome back, operator. The reactor is stable, but I would not call it safe.
• In this episode: why the smallest sound in a scene is usually the one you notice.
• Objective updated. Reach the north gate before the storm closes the pass.

Your script text, a voice from the catalogue, a language, and four delivery controls: stability, similarity, style and speed. Each voice carries a preview you can hear before choosing it.

One audio file per run, as WAV or MP3 at the sample rate you choose. Longer scripts come back as one continuous read rather than a file per sentence.

WAV, FLAC and MP3, at 44.1 or 48 kHz and 16 or 24-bit. You pick the container and the rate on the form before you run it.

Game developers who need a thousand lines of placeholder dialogue this week and cannot book a booth for it. Video creators narrating a cut without recording. Teams localising a script into a language nobody in the room speaks.

One credit per two hundred characters. A five-hundred-word script is about fifteen credits, and the estimate is shown before the run.

Every run is a queued job. It holds its credit estimate while it works and settles to the real cost when it finishes; if it fails or you cancel it, the hold is released in full. You see the estimate before you spend anything.

See pricing

Everything you generate is royalty-free for commercial use — client work, monetised video, games you sell, ads. No per-use fee and no attribution requirement. Generations are yours; uploads stay private by default, and the inputs to the isolator and splitter are deleted automatically.

Which languages are supported?

The voice catalogue is filterable by language, and what is available depends on the engine serving the capability. The filter is on the form, so you can see which languages have voices before you write a script for one.

What do stability and similarity actually change?

Stability trades consistency against expression: a low value performs the line harder and varies more between runs, a high value reads it evenly. Similarity controls how closely the output tracks the chosen voice rather than drifting toward a generic read.

Can I use the speech in a commercial game or video?

Yes. Generated audio is royalty-free for commercial use with no attribution requirement. If you need a specific person’s voice, that is a rights question about the voice, not about this tool.