Free text-to-speech (TTS) or speech synthesizers

Your computer almost certainly comes with at least one synthesizer. Later versions of Windows have better voices and more of them. Below are some free options that all support SAPI5 - that is, they will work with most software on a Microsoft Windows machine. All will work with the Thunder and NVDA free screenreaders.

Universities and public bodies regularly produce new speech synthesizers from research projects and attempts to support their language or accent. These either disappear from the web after a year or two or get brought up and commercialised by a new or existing speech synthesis company. eSpeak is the exception: it's been around for a decade at least. But any of the other links below may be broken or no longer provide the voice described.

Microsoft Windows

Microsoft Windows has shipped with a speech synthesizer since Windows 2000 - Microsoft Sam until Microsoft Vista, other voices since then. English and Chinese were usually freely available. Windows 8 added many speech synthesizers available for lots of languages. You simply install the "Language Pack" from Control Panel and the voice will appear.

Confusingly, Windows 10 also has many voices that are Windows Runtime voices or Mobile voices, not the SAPI5 voices that will work in your software. See Microsoft Speech on the Blog for more details. None of the below are these Windows Runtime voices.

All these SAPI5 voices can be expected to be found on Windows 8.1 and 10, either because (1) I've installed a language pack and found them or (2) they are on the Microsoft website, though incorrectly labeled.

Windows 8 SAPI5 Voices
Korean Heami
US English David
UK English Hazel
US English Zira
Spanish Helena
French Hortense
German Hedda
Japanese Haruka
Chinese (simplified) Hanhan and Huihui
Windows 8.1 and 10 SAPI5 Voices - all the above plus:
Mexican Spanish Sabina
Indian English Heera
Chinese (traditional) Tracy
Italian Elsa
Polish Paulina
Portuguese Maria (possibly Brazilian)
Russian Irina

Here are the Windows Mobile voices for Windows 10 Creators Edition (Spring 2017). Original Microsoft list. Your assistive tech may or may not be able to use them (Narrator can now).

Windows 10 Mobile Voices
Arabic Hoda & Naayf
Brazilian Portuguese Daniel & Maria
Bulgarian Ivan
Catalan Herena
Chinese (Simplified) Kangkang, Huihui & Yaoyao
Chinese (Traditional, Hong Kong SAR) Danny & Tracy
Chinese (Traditional, Taiwan) Zhiwei, Yating & Hanhan
Croatian Matej
Czech (Czech Republic) Vit
Danish Helle
Dutch Frank
English (Australia) James & Catherine
English (Canada) Richard & Linda
English (Great Britain) George, Hazel & Susan
English (India) Ravi & Heera
English (Ireland) Shaun
English (United States) David, Mark & Zira
Finnish Heidi
Flemish (Belgian Dutch) Bart
French (Canada) Claude & Caroline
French (France) Paul, Hortence & Julie
German (Germany) Stefan, Hedda & Katja
German (Switzerland) Karsten
Greek Stefanos
Hebrew Asaf
Hindi Hemant & Kalpana
Hungarian Szabolcs
Indonesian Andika
Italian Cosimo & Elsa
Japanese Ichiro, Ayumi & Haruka
Malay Rizwan
Norwegian Jon
Polish Adam & Paulina
Portuguese (Portugal) Helia
Romanian Andrei
Russian Pavel & Irina
Slovak Filip
Slovenian Lado
Korean Heami
Spanish (Spain) Pablo, Helena & Laura
Spanish (Mexico) Raul & Sabina
Swedish Bengt
Tamil Valluvar
Thai Pattara
Turkish Tolga
Vietnamese An

eSpeak

eSpeak is an open-source (GPL V3) TTS system. There's a SAPI5 DLL, and commandline options. It's developed by Jonathan Duddington. It's probably the most common speech synthesizer in open-source systems at the time of writing. It sounds like a "traditional" speech synthesizer - quite robotic - but this means it's exceptionally clear, high-performance and easy to add language support to it. I know many blind people who use it exclusively in preference to the commercial "human-sounding" voices.

You get:

On Windows, you can install eSpeak then run "C:\Program Files (x86)\eSpeak\command_line\espeak.exe". The commandline options:

Asharir - Hindi, Tamil, Kannada and Telugu

These were free SAPI5 text-to-speech voices from Asharir in India. However, they've disappeared from the Internet, so until someone complains a download link follows. They supported Hindi, English, Tamil, Kannada, and Telugu, though I can't vouch for them.

Welsh

Maltese

Ekho - Chinese, Tibetan, Korean