Estimated reading time at 200 wpm: 7 minutes
As some may not be aware, I’ve been developing my own software to convert text into high quality electronic speech. That development continued for several hours today. As I will remind myself – as if I need to – £490 for 1.6 hours for human narration vs AI-driven production estimated at ~£1.51 for 100 minutes is a freaking good deal and quality will be almost indistinguishable from human narration. I have two books coming following on The Kael’Nyrin Scrolls: Our Children of Starlight. So I need to get ready for audiobook production of each.
Whether or not you agree our Fat Disclaimer applies
So – today’s software development was good. It turns out that Google TTS using Gemini 2.5 pro, is not quite efficient. Instead Google’s Cloud TTS is better: a) faster and b) provides British accents via Chirp. Don’t ask me what Chirp is – I don’t know – don’t need to know – and don’t care! But everybody has choice – so they can choose to ‘understand everything’ completely before putting one foot in front of the next! Do I give a flying flamingo? I do not!
Well, that meant that my Python app, had to be tweaked to work with Cloud TTS. I don’t know Python – cannot write a line of code in Python – and don’t care to learn Python either! Two AIs were roped in to do the work, and I orchestrated development on the nitty gritty! In the end the App was created and it worked perfectly! As I always say, “The Pilot does not need to know exactly how the plane’s engines work, in order to fly the plane!” Of course, being NOBODY – nobody needs to believe me.
So what that means is that for chapters to be turned into audio, I have a streamlined, efficient and cost effective process. Hear or not the sample below for a few words from Alice in Wonderland.
Phonemes
This not a tutorial so I’m not explaining ‘all about it’. Basically there are several words in British English that a pronounced differently depending on context. So bits of code needed to be inserted for the TTS to know how to pronounce them. The table below is for me to use as I go on.
Heteronyms (same spelling, different pronunciation)
| Word | Meaning / distinction | SSML (copy directly) |
|---|---|---|
| bass | instrument | <phoneme alphabet="ipa" ph="beɪs">bass</phoneme> |
| bass | fish | <phoneme alphabet="ipa" ph="bæs">bass</phoneme> |
| bow | weapon / ribbon | <phoneme alphabet="ipa" ph="bəʊ">bow</phoneme> |
| bow | bend | <phoneme alphabet="ipa" ph="baʊ">bow</phoneme> |
| close | near (adj) | <phoneme alphabet="ipa" ph="kləʊs">close</phoneme> |
| close | shut (verb) | <phoneme alphabet="ipa" ph="kləʊz">close</phoneme> |
| conduct | noun | <phoneme alphabet="ipa" ph="ˈkɒndʌkt">conduct</phoneme> |
| conduct | verb | <phoneme alphabet="ipa" ph="kənˈdʌkt">conduct</phoneme> |
| content | noun | <phoneme alphabet="ipa" ph="ˈkɒntɛnt">content</phoneme> |
| content | adjective | <phoneme alphabet="ipa" ph="kənˈtɛnt">content</phoneme> |
| dove | bird | <phoneme alphabet="ipa" ph="dʌv">dove</phoneme> |
| dove | past of dive | <phoneme alphabet="ipa" ph="dəʊv">dove</phoneme> |
| lead | to guide | <phoneme alphabet="ipa" ph="liːd">lead</phoneme> |
| lead | metal | <phoneme alphabet="ipa" ph="lɛd">lead</phoneme> |
| live | verb | <phoneme alphabet="ipa" ph="lɪv">live</phoneme> |
| live | adjective | <phoneme alphabet="ipa" ph="laɪv">live</phoneme> |
| minute | time | <phoneme alphabet="ipa" ph="ˈmɪnɪt">minute</phoneme> |
| minute | very small | <phoneme alphabet="ipa" ph="maɪˈnjuːt">minute</phoneme> |
| object | noun | <phoneme alphabet="ipa" ph="ˈɒbdʒɪkt">object</phoneme> |
| object | verb | <phoneme alphabet="ipa" ph="əbˈdʒɛkt">object</phoneme> |
| permit | noun | <phoneme alphabet="ipa" ph="ˈpɜːmɪt">permit</phoneme> |
| permit | verb | <phoneme alphabet="ipa" ph="pəˈmɪt">permit</phoneme> |
| present | noun | <phoneme alphabet="ipa" ph="ˈprɛzənt">present</phoneme> |
| present | verb | <phoneme alphabet="ipa" ph="prɪˈzɛnt">present</phoneme> |
| produce | noun | <phoneme alphabet="ipa" ph="ˈprɒdjuːs">produce</phoneme> |
| produce | verb | <phoneme alphabet="ipa" ph="prəˈdjuːs">produce</phoneme> |
| read | present | <phoneme alphabet="ipa" ph="ɹiːd">read</phoneme> |
| read | past | <phoneme alphabet="ipa" ph="ɹɛd">read</phoneme> |
| record | noun | <phoneme alphabet="ipa" ph="ˈrɛkɔːd">record</phoneme> |
| record | verb | <phoneme alphabet="ipa" ph="rɪˈkɔːd">record</phoneme> |
| refuse | noun | <phoneme alphabet="ipa" ph="ˈrɛfjuːs">refuse</phoneme> |
| refuse | verb | <phoneme alphabet="ipa" ph="rɪˈfjuːz">refuse</phoneme> |
| row | line | <phoneme alphabet="ipa" ph="rəʊ">row</phoneme> |
| row | argument | <phoneme alphabet="ipa" ph="raʊ">row</phoneme> |
| tear | from eye | <phoneme alphabet="ipa" ph="tɪə">tear</phoneme> |
| tear | rip | <phoneme alphabet="ipa" ph="tɛə">tear</phoneme> |
| use | noun | <phoneme alphabet="ipa" ph="juːs">use</phoneme> |
| use | verb | <phoneme alphabet="ipa" ph="juːz">use</phoneme> |
| wind | air | <phoneme alphabet="ipa" ph="wɪnd">wind</phoneme> |
| wind | twist | <phoneme alphabet="ipa" ph="waɪnd">wind</phoneme> |
| wound | injury | <phoneme alphabet="ipa" ph="wuːnd">wound</phoneme> |
| wound | past of wind | <phoneme alphabet="ipa" ph="waʊnd">wound</phoneme> |
Homophones and function‑word checks (included deliberately)
These normally do not require SSML, but are included so you can force pronunciation if a voice/model behaves oddly.
| Word | Word | Word |
|---|---|---|
| for | /fɔː/ | <phoneme alphabet="ipa" ph="fɔː">for</phoneme> |
| four | /fɔː/ | <phoneme alphabet="ipa" ph="fɔː">four</phoneme> |
| to | /tuː/ | <phoneme alphabet="ipa" ph="tuː">to</phoneme> |
| too | /tuː/ | <phoneme alphabet="ipa" ph="tuː">too</phoneme> |
| two | /tuː/ | <phoneme alphabet="ipa" ph="tuː">two</phoneme> |
| by | /baɪ/ | <phoneme alphabet="ipa" ph="baɪ">by</phoneme> |
| buy | /baɪ/ | <phoneme alphabet="ipa" ph="baɪ">buy</phoneme> |
| here | /hɪə/ | <phoneme alphabet="ipa" ph="hɪə">here</phoneme> |
| hear | /hɪə/ | <phoneme alphabet="ipa" ph="hɪə">hear</phoneme> |
| no | /nəʊ/ | <phoneme alphabet="ipa" ph="nəʊ">no</phoneme> |
| know | /nəʊ/ | <phoneme alphabet="ipa" ph="nəʊ">know</phoneme> |
| one | /wʌn/ | <phoneme alphabet="ipa" ph="wʌn">one</phoneme> |
| won | /wʌn/ | <phoneme alphabet="ipa" ph="wʌn">won</phoneme> |
| right | /raɪt/ | <phoneme alphabet="ipa" ph="raɪt">right</phoneme> |
| write | /raɪt/ | <phoneme alphabet="ipa" ph="raɪt">write</phoneme> |
| their | /ðɛə/ | <phoneme alphabet="ipa" ph="ðɛə">their</phoneme> |
| there | /ðɛə/ | <phoneme alphabet="ipa" ph="ðɛə">there</phoneme> |
| they’re | /ðɛə/ | <phoneme alphabet="ipa" ph="ðɛə">they’re</phoneme> |
| your | /jɔː/ | <phoneme alphabet="ipa" ph="jɔː">your</phoneme> |
| you’re | /jɔː/ | <phoneme alphabet="ipa" ph="jɔː">you’re</phoneme> |
Conclusions
- The ability to read and understand is a good thing.
- Similarly the ability to follow instructions is also a good thing.
- Persistence pays off.
