live instance · rev 0.31

ECHOES
your digital
clone.

Train Echoes on your voice, chats, and memories. It can reply across WhatsApp, voice notes, calls, and group chats with your tone intact.

create account→ sign in→
i. the brain knowledge graph

Memories become a digital brain.

Every message runs hybrid retrieval — vector search alongside keyword match — pulling the right context in real time, so the clone always sounds like you today, not a snapshot from last month.

node types · 9
memory—
relationship—
trait—
speech—
opinion—
belief—
emotion—
routine—
preference—
ii. iron interface hand & voice control

Explore it with your hands.

Iron Mode opens webcam hand tracking with pinch to zoom, sweep to orbit, and voice running in parallel — the brain arrives like a planetarium you can reach into.

pinch right hand
Zoom into the brain. Camera holds still the instant fingers separate — no drift.
sweep left hand
Orbit in the same direction your palm moves. Active filter re-targets the centroid.
pinky out left hand
Toggles the inspector panel with matching memory cards.
voice parallel
"search", "filter relationships", "tell me about …", "rotate", "add memory", "music lower".
wake word voice
"brain", "wake up" — TTS welcomes you, ambient music fades in. "sleep" exits.
iii. whatsapp always on

Bring your clone into WhatsApp.

Not just notifications. Real conversations, voice notes, group replies, and relationship-aware context from the same memory core.

It can sound present — not automated.

Echoes can keep up inside WhatsApp with your tone, your history, and the right relationship context, so the clone feels like an extension of you instead of another dead bot layer.

i. direct
1:1 conversations
Reply naturally in private chats with long-term memory, recent context, and the same voice across every exchange.
ii. groups
Group-aware replies
Track who said what, which group it is, and how you relate to each person — so answers land with the right social context.
iii. voice
Voice notes built in
Send spoken replies with your cloned voice when text is too flat, too slow, or just not the right move.
iv. capabilities what ships

Memory, channels, presence — one engine.

The product is a usable digital clone, not a static memorial. Memory, voice, messaging rails, and live context all ship together — talk to your clone, or let it reply for you.

i.core
Knowledge graph
A 12-table schema across stories, traits, relationships, speech, opinions, emotions, routines, preferences, and beliefs.
ii.retrieval
Hybrid RAG
768-dim vector search plus keyword match on every turn, with a fallback to high-importance recent memories.
iii.presence
Scan-line particle face
Portraits become 24,000+ monochrome particles. Audio frequency drives jaw, brows, and mouth; the body breathes and sways.
iv.voice
Voice cloning
Samples accumulate into one neural voice model. A quality meter scales from minimal to premium as more audio arrives.
v.groups
Context-aware replies
Understands who is speaking, which group it is in, and the relationship context — so replies land like you, not like a blind bot.
vi.voice
Voice notes & calling
Send voice messages, answer with cloned speech, and place live calls when text is not enough — backed by your memory and tone.
vii.avatar
Video calibration
A short phoneme-complete recording builds a viseme bank and motion pool for dramatically tighter avatar lip-sync.
viii.identity
Voice-based ID
256-dim speaker embeddings recognize who is talking and shift context to that relationship in real time.
v. pipeline upload to conversation

Five steps to a living memory.

Add what you have. Echoes does the structuring.

i.
Capture
Type, dictate, drop in files, or run the 201-question guided interview. The voice mic is on every input.
ii.
Extract
AI parses everything into typed entities with confidence scores. High commits; mid is queued for your review.
iii.
Embed
Each entity becomes a 768-dim vector and joins the knowledge graph. Per-person snapshots back up the entire brain before every import.
iv.
Embody
Optional: clone the voice, scan the face into a particle cloud, calibrate from a short reading video for sharper lip-sync.
v.
Talk
Open chat. Tap the mic. Speak. The hologram answers in your voice, in your language, with your memories — in real time.

Deploy your digital clone.

Train an emulator of you that replies across socials, WhatsApp, group chats, voice notes, and calls — without losing your memory or your tone.

create account→ sign in→