eidosSpeech v3 is Here: Real SSML, Long-Text & New Free Tools
v3 is our biggest update yet — and our most honest. We fixed features that never really worked, shipped genuinely new ones, and hardened the whole platform. Here's everything that changed, plus a full changelog.
Since the last release we've done a deep pass over the entire product: what works, what only pretended to work, and what would genuinely help you. The result is eidosSpeech v3. It stays 100% free, no credit card, no sign-up for basic use — just better.
✨ Real SSML control (that actually works)
Previously, "voice styles" tried to use Azure emotion markup that the free voice engine doesn't support — it could even read the markup out loud. In v3 we rebuilt this properly. Expressive styles now map to real, working prosody, and we added a genuine SSML engine that supports the tags the engine can actually honor:
<break>— insert real pauses of any length<prosody>— adjust rate, pitch, and volume per span<emphasis>— make words stand out<sub>— control pronunciation of names, brands, and numbers
Try it in the new, free SSML Playground — write markup and hear the result instantly. New to SSML? Read the SSML guide.
📜 Long-text to speech
The single-request character limit was the most common source of friction. v3 adds a Long Text tool: paste an entire article, chapter, or script and eidosSpeech automatically splits it at sentence boundaries, synthesizes each part, and stitches everything into one downloadable MP3. It understands both space-separated and character-based (CJK) languages.
📱 Install eidosSpeech as an app
eidosSpeech is now a Progressive Web App. Add it to your phone or desktop home screen and it opens like a native app, loads instantly, and even works offline for the parts that don't need the network. No app store required.
🔎 Voice search & history
Finding the right voice among 1,200+ options is now easy: a new search box filters voices by name, language, or gender, on top of the existing language and favorites filters. And a new generation history keeps your recent creations one click away so you can tweak and re-run without retyping.
🌍 Per-language pages
We added dedicated pages for popular languages — Indonesian, English, Japanese, Spanish, French, German, Korean, Chinese, Hindi, and Arabic — each with the right voices and a quick demo, so it's faster to get started in your language.
🔒 Honesty & hardening
We believe in shipping what actually works. In v3 we also:
- Marked experimental voice cloning clearly as coming soon instead of faking results.
- Fixed several input-validation and security edge cases across the API.
- Tightened how requests are bounded so the service stays fast and reliable under load.
📋 Full changelog
Added
- SSML engine with
break,prosody,emphasis, andsubsupport (POST /api/v1/tts/ssml). - SSML Playground web tool with live preview and history.
- Long-text to speech with automatic sentence-aware chunking (
POST /api/v1/tts/long) and a dedicated web tool. - Installable PWA with offline app shell.
- Voice search and generation history in the TTS app.
- Per-language landing pages for 10 languages.
- Three new guides on the blog (YouTube voiceovers, accessibility, SSML).
Changed / Fixed
- Rebuilt voice styles to use real, working prosody instead of unsupported emotion markup.
- Voice cloning is now clearly experimental and returns an honest status instead of fake results.
- Bounded request text length, tightened origin checks, and fixed an authentication edge case.
- Resolved a memory-leak path in voice previews.
- Rebranded the product to v3 across the site.
Try v3 now
Everything is live and free. Open the TTS app, paste something into the Long Text tool, or experiment in the SSML Playground. Developers can explore the new endpoints in the API docs. Thanks for using eidosSpeech — more coming soon.