Category Archives: The Animation Industry

It seems rather odd to consider old-school text-to-speech software and SAPI5 voices, at a time when Poland’s ElevenLabs is doing such great things with AI-generated voices. But I’m always one to cherish old Windows freeware, and at present all the new AI voices are online and require a monthly/yearly subscription. So I was pleased to find an alternative freeware to Balbolka for desktop PC text-to-speech using SAPI5 voices. Many such voices are also now abandonware on Archive.org, the key companies having since been sold on several times.

Made in Italy, the DSpeech TTS freeware used to be fairly basic, but it’s improved enormously since about 2016. Though this is not a fact reflected in its rather basic 1990s-style download page, which you’ll have to overlook. This freeware is now in version 1.74 (spring 2022), It’s genuine one-man freeware, made in Italy, and is feature-comparable with Balbolka though a bit rougher in UI and Help translation to English.

The DSpeech download link uses only a .GIF button, so if you have a .GIF blocker in your Web browser, then you instead right-click the page and ‘View Source’. You should then see a live working link to the download in the HTML…

The English manual is included in the software. There’s no Windows installer, just unzip where you want and run it.

SCRIPTING: Beyond the usual control tags, DSpeech supports basic scripting including voice-recognition and script loops. Which is unusual. Apparently it can even read out VLCplayer movie sub-titles in real-time, in a chosen SAPI5 TTS voice and speed.

TAGS: The tagging menus make switching voices easy. There’s better right-click support than in the latest Balabolka for adding tags, though that’s not saying much. When you highlight a word in DSpeech, and add a tag, the word is not wrapped with a closing and opening tag, it’s deleted. Urgh! Having right-click is great, but… the rest of the tag insertion system is not good.

LOQUENDO: DSpeech is supposed to support Loquendo ‘voice expressions’ (laugh, sigh etc) via the Italian Loquendo 6 Italian ‘Paola’ and ‘Luca’ TTS voices, combining words with special expressive tags such as \_Laugh and suchlike. Later the tag syntax was changed to \item=Laugh in Loquendo version 7 voices. But while these v6 voices work fine in any DSpeech, and v7 voices work fine in DSpeech v.1.72.29 (December 2018, not the latest 1.74.x), their expressive cues no longer vocalise in DSpeech. You just hear silence.

Spanish Loquendo 7 voices (not 6) can however ‘express’ when used in Loquendo’s own Java-based TTS Director, which came with the Loquendo SDK. See YouTube for examples and useful links.

Regrettably neither the Loquendo 6 or 7 voices can even be played in the other TTS freeware Balabolka, though they do show up on its voice menu. It thus seems that properly-working Loquendo voices are limited to…

* Loquendo 6 (any voice) on DSpeech 1.74.x or earlier. Loquendo 7 not supported on the latest DSpeech.

* Loquendo 7 (Spanish) on DSpeech 1.72.29 (or earlier?), or Loquendo 7 (Spanish) on Loquendo TTS Director with SDK and Spanish pack.

The Spanish version 7 voices do however have ‘expressives’ that work fine with Loquendo TTS Director 7, which was Windows freeware which shipped with the developer/API/SDK kit. This success at least showed me that the problem was not with my PC or a 32-bit / 64-bit Windows clash, at least for version 7 voices.

Yet it’s strange. Obviously DSpeech could, at one time, play the ‘expressives’ in the Loquendo 6 voices. But, no longer, it seems. Switching back to an older DSpeech 1.72.29 didn’t cure the problem, but it did usefully fix the playing of the Loquendo 7 voices. I suspect that Loquendo 6 voices now have a 32-bit / 64-bit problem on 64-bit Windows, despite the player and voices both being 32-bit.

Loquendo TTS Director voices have a complete list of expressives in the C:\Program Files (x86)\Loquendo\LTTS7\data\voices\Soledad\SoledadGildedParalinguistics.sde file (change name for each voice). Open it in Notepad++ to see the list in plain-text. For instance, Soledad has the following, and obviously you can also mix and match and tone-shift…

\item=Ah
\item=Ah_01
\item=Ah_02
\item=Ay
\item=Ay_01
\item=Breath
\item=Breath_01
\item=Buh
\item=Buh_01
\item=Buuu
\item=Buuu_01
\item=Cataplum
\item=Cataplum_01
\item=Click
\item=Click_01
\item=Click_02
\item=Cough
\item=Cough_01
\item=Cough_02
\item=Cry
\item=Cry_01
\item=Cry_02
\item=Cry_03
\item=Ehm
\item=Ehm_01
\item=Epa
\item=Epa_01
\item=Hey
\item=Hey_01
\item=Hiccup
\item=Hiccup_01
\item=Hiccup_02
\item=Laugh
\item=Laugh_01
\item=Laugh_02
\item=Mhmm
\item=Mhmm_01
\item=Mhmm_02
\item=Mhmm_03
\item=Oh
\item=Oh_01
\item=Ohoh
\item=Ohoh_01
\item=Ops
\item=Ops_01
\item=Prrr
\item=Prrr_01
\item=Shhh
\item=Shhh_01
\item=Sigh
\item=Sigh_01
\item=Sigh_02
\item=Singing
\item=Singing_01
\item=Smack
\item=Smack_01
\item=Smack_02
\item=Sniff
\item=Sniff_01
\item=Sniff_02
\item=Sniff_03
\item=Snore
\item=Snore_01
\item=Swallow
\item=Swallow_01
\item=Throat
\item=Throat_01
\item=Throat_02
\item=Throat_03
\item=Uff
\item=Uff_01
\item=Ups
\item=Ups_01
\item=Whistle
\item=Whistle_01
\item=Whistle_02
\item=Whistle_03
\item=Whistle_04
\item=Whistle_05
\item=Yawn
\item=Yawn_01
\item=Yawn_02
\item=Yeee
\item=Yeee_01
\item=Yuhu
\item=Yuhu_01
\item=Yuhu_02
\item=Zas
\item=Zas_01

Easier to just paste these all in and cut out what you don’t want. Rather than wrestling with menu-based insertion.

VOICEWARE: DSpeech has support for reading with a VoiceWare TTS, but not for a vital aspect of the voice. The first version of the TTS VoiceWare voices (e.g. VW Bridget, British) had different inflections on words if you added ! ?! or !? (again, see YouTube for demo and useful links). But this feature of the voice is not supported in DSpeech. It is supported in Balabolka. So this is another deal-breaker for DSpeech.

CONCLUSIONS: Despite what at first glance seems to be DSpeech’s more intuitive right-click tag adding, Balabolka is on several counts the superior tool for longer-form editing. It properly wraps highlighted words in starting/closing tags, which is vital if you’re TTS-coding something longer than a paragraph. It also supports VoiceWare’s ! ?! and !?, useful for one of the best British voices.

I thus suggest using the latest Balabolka for freeware TTS scripting and recording, and the old Loquendo TTS Director + its Spanish voices for creation of vocal FX, pitch and speed-shifted to match the voice being used in Balabolka. Then embed these vocal FX as audio clips in Balabolka. This is not as ideal as having Balabolka support Loquendo (it refuses to even read their voices), but it’s a viable workaround.

The ideal would be to have a standard SAPI5 voice that was ‘expressives only’, for use in Balabolka. A sort of audio FX bank, that could be reliably called with a simple tag (such as \_sneeze etc). But so far as I can see, that doesn’t exist, other than by chopping bits from my Dictionary of British Pronunciation for TTS.

Finally, note that TTS Director only ‘sees’ its own Loquendo voices, and is therefore no good as a general SAPI5 TTS script editor. TTS can be done in Adode Captivate (used for super-Powerpoint ‘e-learning’ creation) and in CrazyTalk / Cartoon Animator, but the editing is not at all comparable to Balaboka.