Text to speech fails on the small things. A Wikipedia sentence turns into "three hearts open bracket twelve close bracket". A novel stops after "Mrs." as if the sentence ended there. A history book reads "Henry VIII" as "Henry vee eye eye eye". In Aloudeo, the voice gets clean text before it speaks.
Here is a real example
A line from a web page:
The common octopus (/ˈɒktəpʊs/ OK-tə-pəs) has three hearts[12] and blue blood,[13] as Fig. 3 shows.
What you hear:
The common octopus has three hearts and blue blood, as Figure 3 shows.
What stays off the voice
- Links, web addresses and email addresses
- Citation brackets like [12], [a] and [citation needed]
- Author-year citations like (Smith & Lee, 2019)
- Footnote daggers and stars stuck to a word
- Page markers like [Pg 23] and editor lines in brackets
- Pronunciation guides like (/ˈɒktəpʊs/)
- Pointers like (see Table 3) or (Figure 2b)
- Emoji, bullets and arrows that point at nothing
What you hear in full
| On the page | What you hear |
|---|---|
| Mr. Bennet | Mister Bennet |
| St. Paul's | Saint Paul's |
| Chapter IV | Chapter four |
| Henry VIII | Henry the eighth |
| World War II | World War two |
| 3/8 cup | three eighths cup |
| 11/15/2026 | November 15, 2026 |
| 24/7 | twenty-four seven |
| CO2 | C O 2 |
| 5 p.m. | five P M |
| e.g. | for example |
Where it stops
The rules cover English text. A name the voice has never seen can still come out oddly, and a page made of photos can only read the text your iPhone recognized on it.