Himaia: The Open Persona Voice API

Himaia is a voice infrastructure platform built for creators, modders, and game developers who need characters that stay true to their identity. Unlike standard Text-to-Speech tools that just provide a static sound, Himaia treats a Persona as the core unit of identity. This ensures that a character's voice, tone, and personality remain consistent across different scenes and contexts. It solves the problem where characters might sound the same whether they are whispering at night or shouting during a fight.

Benefits

Himaia offers several key advantages over traditional voice APIs. First, it ensures consistency by keeping a character's identity intact regardless of the emotional context or scene. Second, it simplifies development by allowing users to define a character's identity in a single file called a persona.yaml file. This means developers can generate audio with just one API call instead of building complex pipelines with multiple components. Third, the platform provides an open specification under Apache-2.0, which allows users to commit personas to Git and migrate runtimes freely. Finally, it offers a wide selection of 30 pre-tuned voices that can be mixed and matched with any persona to create unique sounds.

Use Cases

Himaia is designed for specific groups of people who need realistic character interactions. Modders and power users can use it with extensions like SillyTavern to make character cards talk without sounding like generic demos. Indie teams can build companion and chat-character apps without maintaining a fragile system of prompts and emotion tags. Game developers can use it for Foundry GMs, Unity solo projects, and Interactive Fiction tools. In these scenarios, the same NPC can be used across every scene with just one persona file, making it easier to ship character applications quickly.

Pricing

Himaia operates on a model that starts free and scales as needed. The free tier costs nothing and includes 20 voiced minutes per month along with access to 3 personas. New accounts receive credits on signup that refresh monthly. For those who need more, the Creator Tier costs $19 per month and offers 300 voiced minutes plus 10 private personas. The Pro Tier is priced at $79 per month and provides 1,000 voiced minutes plus 100 cinematic minutes. There are also specific rates for usage including basic lines at $0.04 per minute and in-character lines at $0.06 per minute. Enterprise customers can contact the team for custom pricing.

Vibes

Users appreciate Himaia for its ability to ship characters without building complex pipelines. The platform is praised for solving the issue of persona drift where characters forget their identity after many interactions. Developers find the single-file setup and open specification very helpful for collaboration and long-term projects. The free tier is particularly well-received because it allows users to test and ship demos immediately without needing a credit card.

Additional Information

Himaia was founded to address the limitations of standard voice APIs by making the Persona the unit of work. The team has released the voice.persona specification as open-source software. They offer starter personas like Dry Butler, Tavern Rogue, and Weary GM to demonstrate the system's versatility. The platform supports various fidelity modes such as verbatim, shape, and rewrite to control how input is processed. Future integrations include a Foundry VTT extension and a Unity package to expand its reach in the game development community.