Animus is a small, deliberate piece of software. This page lists the services and tools behind it, what they do, and what we choose not to do. Transparency is part of the contract.
Everything Animus says and sounds like passes through one of two outside services. We chose them deliberately, and we don't share your data with anyone else.
Reading what you wrote, finding the right memory for the right question, transcribing what you say.
gpt-4o
Conversations and the family-side chat.
gpt-4o-mini
Memory titling, persona overview, auto text-tuning.
whisper-1
Transcription when you talk to Animus.
text-embedding-3-small
Finding the right memory by meaning, not keyword.
Learning the way you sound, then speaking in your voice when invited family members visit.
Instant Voice Cloning
Builds and re-trains your voice model from short samples.
eleven_multilingual_v2
The text-to-speech model that speaks in your voice.
mp3 192kbps
Optional higher-quality audio (Creator plan or above).
Animus is intentionally simple. No databases, no clusters, no analytics SDKs. Files on disk, served by one process.
One process serves both the website and the API. Sessions live in cookies. Sensitive routes are gated server-side, not client-side.
No framework, no build step. What you see in the browser is what's in the file. Easier to read, easier to audit, easier to keep small.
Memories, persona settings, conversations, voice samples — each stored as plain JSON or media in a per-user folder. No relational database.
A short list. Each one is open-source and credited where it matters in the codebase.
^5
openai ^6
axios ^1
multer ^2
bcrypt ^6
cors ^2
dotenv ^17
nodemailer ^8
form-data
A short ledger. The longer story is in the code itself.