Arabic Text to Speech With Emotion: How to Get Excitement, Calm or Sorrow

01Where emotion in a synthetic voice comes from

Plenty of people try text to speech once, type an excited line, hear it read like a news bulletin and decide AI "has no feeling". Modern models do read context: is this a question, an exclamation, a condolence? But they work with what you give them. A flat script, a formal voice and neutral settings produce a neutral read.

Think of it like directing a voice actor: you cast the right person, give them a script written for the ear, and tell them how you want it delivered. In Vocla those are five levers, from most to least influential:

02Cast a voice that already carries the mood

The biggest emotional decision is the voice itself. A deep, calm voice won't become a hyped youth ad no matter how far you push the sliders, and a bright, fast voice won't suit a condolence message. Start in the Voice library: filter by dialect and gender, then audition each voice on a line from your script, not the default preview.

• Ads and promos: youthful, energetic voices, usually in the audience's dialect.
• Meditation and bedtime stories: warm, low voices that are naturally slow.
• Condolences and sensitive notices: serious, quiet voices with no promotional edge.
• Teaching and explainers: clear, balanced voices, neither hyped nor monotone.

If the library doesn't have the voice in your head, use Voice design. Describe it, for example "a Saudi woman in her thirties, warm and calm, for meditation content", and Vocla generates three original candidates; save your favourite to your workspace. You can also add up to five library voices as style references when you're close but not quite there.

03Wording and punctuation: write for the ear

Models pick up emotion from words more than people expect. "We are pleased to announce product availability" will never sound excited, because it isn't an excited sentence. "Finally! The one you've been waiting for is here" lifts the energy with almost any suitable voice. Write the way a person who feels that emotion would actually speak. In Arabic, that often means writing in the dialect rather than formal MSA:

Punctuation is the simplest form of direction. These rules hold for most voices:

04Vocla's voice settings: what each slider does

In Text to speech, the Voice adjustment panel has five sliders, plus an Energy mode with two options. Here's what each does, as labelled in the app:

Consistency and expressiveness together are the key. High consistency with low expressiveness gives you a steady newsreader. Medium consistency with high expressiveness gives a lively read with real rise and fall. Push both to the extremes and the delivery can vary unpredictably between generations. Save settings you like as a preset.

The Delivery panel also has an Extra instructions field: describe the read in your own words, for example "read this in a sad, slow tone and pause after each sentence", and it applies to that text only. Some models read text exactly as written and ignore delivery notes; the app tells you when that's the case and suggests switching models.

05Ready examples for each mood: hype, calm, condolence

Three scripts with suggested settings. Try them as they are, then change one setting at a time to learn what it does.

Settings: a youthful, energetic voice, Energy mode Boost, expressiveness above the middle, consistency in the middle, speed slightly up. Extra instructions, if the model supports them: "excited and happy, like telling a friend great news".

Settings: a warm, low voice, Energy mode Normal, speed below the middle, high consistency, expressiveness near neutral. The short, full-stopped sentences do half the work.

Settings: a serious, quiet voice, Energy mode Normal, speed slightly down, high consistency, low expressiveness. No upbeat music, and check the vowelling of names and Quranic phrases with the tashkeel tool before generating.

06Dialogue with a different emotion on every line

For a conversation, such as a radio ad or a short scene, use Text to dialogue. Write the script as "Name: line", give each speaker a voice, then choose one of seven delivery styles per line: natural, excited, calm, whisper, serious, happy or sad. Add a pause after any line, and Vocla mixes everything into one file with the loudness evened out.

Give Sara's first line "excited", Fahad's "natural" with the surprise coming from the words, and the last line "happy". Variety between lines is what makes a dialogue sound real rather than stitched together.

07Limits worth knowing

• Extreme emotion is hard: real sobbing, screaming or long laughter rarely comes out fully convincing from any synthetic voice.
• Results vary by voice: the same settings sound different on every voice, so test on the one you'll actually use.
• Generations aren't always identical: with low consistency the read can shift slightly; generate twice and keep the best.
• Dialect shapes feeling: Saudi excitement isn't Egyptian excitement, so write in the voice's dialect rather than in MSA.

The golden rule: change one setting at a time and listen. Change voice, speed and expressiveness together and you'll never know which one made the difference. When you land on a sound you like, save it as a preset with a clear name such as "Saudi hype ad" or "Calm meditation".