SpeakScribe has 75 English voices: 28 Studio voices that run on our own servers and 47 Global Microsoft neural voices. They cover 14 accents, from American and British to Indian, Nigerian, Kenyan and Philippine English, so start by choosing the accent your viewers expect.
How English is written — and what to type so the voice reads it right
Spelling does not change the accent. Typing 'colour' will not make an American voice sound British, so pick the voice that matches your audience.
Heteronyms are the main source of wrong readings: words such as read, lead, live, record, wind and tear are spelled the same but pronounced differently. The voice guesses from context, so rephrase the sentence rather than respelling the word.
Replace 'I have read it' with 'I finished reading it' when the past tense is meant.
Dates, abbreviations and acronyms in English scripts
Numeric dates are the most common error. 03/04/2026 means 3 April in the UK and 4 March in the US, so write the month name.
Abbreviations are ambiguous: 'St.' can be Street or Saint, and 'Dr.' can be Doctor or Drive. Write the full word when it matters, and preview any acronym that could be read either as letters or as a word.
UK order: '4 March 2026'. US order: 'March 4, 2026'. Match the order to your audience.
Accents and regional varieties
English varieties differ in vowels, rhythm and vocabulary, and every voice is labelled with its accent.
No voice is labelled Scottish, Welsh, Jamaican or other Caribbean English. For those audiences, choose the closest British or American voice and listen to a test paragraph before you publish.
Choosing between the English voices
The 11 American female Studio voices differ most in pace and intonation. Nicole is the slowest reader of all 75 voices measured (118 words per minute) and has the widest intonation range in the group (17.8 semitones), which suits calm, expressive narration. River reads fastest in the group at 228 words per minute with a steady 5.6-semitone range. For a lower voice, Alloy (146 Hz) and Kore (152 Hz) sit at the bottom of the group; Jessica (208 Hz) and Bella (203 Hz) sit at the top.
Among the 9 American male Studio voices, Onyx is the deepest at 86 Hz and Eric the highest at 156 Hz, as well as the fastest reader in the library (230 words per minute). On the British side, Lewis is the deepest voice of all 75 (85 Hz), Fable has the widest intonation range in the library (24.1 semitones), and George is the highest British male Studio voice at 140 Hz with an even 180 words per minute.
Global voices keep their natural pace, so choose them by accent first. For US audiences, Ava (201 Hz) and Andrew (112 Hz) are a clear female/male pair; Guy (155 Hz) is the highest American male Global voice and Eric (106 Hz) the lowest. For India, Neerja (258 Hz) sits far above Prabhat (121 Hz). For Nigeria, Ezinne (192 Hz) and Abeo (115 Hz) give the same female/male contrast. Maisie (British, 299 Hz) and Ana (American, 296 Hz) are the two highest-pitched voices, which fits their use for younger characters.
Common mistakes when turning English text into speech
Writing 'us' for the country: 'US' is the United States, while 'us' is a pronoun. Capitalise the country.
Dropping the comma in direct address: 'Let's eat Grandma' and 'Let's eat, Grandma' mean different things. Add the comma so the pause is in the right place.
Using all capitals for emphasis: a word in capitals may be read letter by letter. Use normal case and punctuation instead.
Mixing 'color' and 'colour' in one script: choose one spelling variety, because mixed spellings also make captions look wrong.
English voices compared — measured
Each voice read the same English text at normal speed. Lower pitch sounds deeper; a wider intonation range sounds livelier, a narrower one steadier.
Measured by SpeakScribe in October 2026 from each voice's audio. Pitch is the median fundamental frequency; intonation range is the spread between the 10th and 90th percentile of pitch. Speaking rate is words per minute on a 59-word passage. st = semitones. Methodology · How SpeakScribe voices work
Which English voice for YouTube narration
A selection of the most-visited English voices on SpeakScribe, ordered by measured intonation range (how much the pitch moves while reading). Across all 75 measured English voices the median is 9.1 semitones and 16.0 characters a second.
-3.4 st vs the median: steadier delivery for tutorials, instructions and calm documentaries.
Same measurements as the table above (October 2026). "Why it fits" follows from the numbers: a range 1.5 semitones or more above the median reads as livelier, 1.5 or more below as steadier; pace is compared with the quarter of voices that read fastest and slowest. Making YouTube videos with SpeakScribe voices →
Global: Libby, Maisie, Sonia, Ryan and Thomas. Studio (speed control, WAV): Alice, Emma, Isabella, Lily, Daniel, Fable, George and Lewis.
Yes. Neerja, an expressive version of Neerja, and Prabhat speak with an Indian English accent. All three are Global voices.
The multilingual Global voices: Ava, Emma, Andrew and Brian (American) and William (Australian). They keep the same voice when your script switches language.
No. The Studio Emma is British and the Global Emma is American, so check the accent label before you choose.
Molly (female) and Mitchell (male) use a New Zealand accent. Emily (female) and Connor (male) use an Irish accent.
Voices pronounce from the spelling, so preview the name and respell it in the script if it is read wrongly.
Yes. Under SpeakScribe's Terms (section 6), audio from the 28 English Studio voices may be used for any purpose, including paid advertising, broadcast and client work. The 47 Global voices may be used in your own content, including monetized videos on your channel; for paid ads, broadcast or client work use a Studio voice. No attribution is needed. Whether a channel earns money is decided by YouTube under its own content policies, which favour original videos.
A free account converts up to 5,000 characters per render, about 5 minutes of English speech at the median pace of 16.0 characters a second. Starter allows 30,000 characters per render (about 31 minutes) and the largest plan 150,000 (about 156 minutes), delivered as one file.
Yes, with Multi-voice: each line gets its own speaker and the result is one MP3 with every voice levelled to the same loudness. Free accounts can try 2 speakers (3 dialogues a month, up to 1,500 characters); paid plans allow 4 to 10 speakers.
Paid plans export SRT and VTT subtitle files with every voiceover, timed from the voice itself; in our test the average offset was 26–97 ms. Free accounts see a preview of the first lines.
Generate English voiceover audio free
Pick a voice above, paste your script, and export publish-ready audio in seconds.