Voices

What it does

Three tiers behind one picker. Four voices live inside the render image and cost nothing to run. Cloud voices bring the accents — American, British, Indian, Australian English; Spanish from Spain and the Americas; Italian, French, Canadian French, German, Brazilian Portuguese, Hindi — thirty named speakers in each. Directed voices are the same thirty speakers reading a line the way you tell them to: "warm, unhurried, a little amused".

A cast member keeps one voice across every episode. Hear a line in any cloud or directed voice before you choose it.

The standard behind it

• A render never fails on a voice. A cloud voice that will not answer is spoken by a local one, with a warning in the log — the line is there, in the wrong voice, which you can hear; silence would say nothing.
• Every line is brought to one sample rate before the narration is joined. Two services at two rates would play one of them fast and short.
• Direction is honoured by the voices that can take it and ignored by the rest, never read aloud as text.