Text to Speech Engine

In Tomodachi Life (3DS), the engine behind the famous “Hillary voice” is Nuance Automotive TTS, specifically an older generation that uses 16 kHz voice models.

It is the engine used in Miitomo, but even though both games share the same TTS engine, Miitomo uses newer models. On Miitomo the models being cfg3 / 3f, while Tomodachi Life its the older cfg1 / 1f models.

The models can be found in the game’s romfs, inside romfs/libNTTS/. The used components, ruleset, and dict resources are stored in their respective .bin files, which are packaged as BARC archives.

How it works

Tomodachi Life does not simply store one prerecorded voice for each Mii. Instead, it generates speech at runtime using the TTS engine and a set of adjustable voice parameters: pitch, speed, quality, tone, accent and intonation. They also seem to prefer female models, probably because they work better with the voice parameters.

This is why Miis can sound very different from each other even though they are all using the same TTS model.

Singing

The TTS system is also tied to singing. Tomodachi Life does not use a completely separate engine for songs; instead, it appears to reuse the speech engine with extra data for timing, effects, and note control.

WIP

Singing still need some documentation, with all the ‘effects’ found on MSBT files.

Differences from Miitomo

Although Tomodachi Life and Miitomo both use Nuance TTS, their resources are not exactly the same.

• Tomodachi Life: older models, commonly described as cfg1 / 1f
• Miitomo: newer models, commonly described as cfg3 / 3f
• Miitomo also uses 22kHz models, while Tomodachi Life still using 16kHz sample rate.

Because of this, voice models from one game may not be directly compatible with the other without additional conversion or patching.

Summary

In short, Tomodachi Life (3DS) uses Nuance Automotive TTS to generate Mii speech in real time. The system relies on older 16 kHz voice models and adjustable synthesis parameters such as pitch, speed, tone, accent, and intonation. It is closely related to the engine later used in Miitomo, but Miitomo uses newer voice models.