Chat With Your
AI Companions
Live inside your browser sidepanel while you browse Grok, Gemini, or anywhere else. No new tab. No separate app. Just open the panel and start talking.
Don't like Grok's voice? Bring your own. Don't trust a cloud LLM? Run it locally. Want a character from a film? Import them from IMDB.
Hey there… didn't expect someone like you to stumble into my quiet corner.
Care to keep a girl company?
This feels like the start of something fun
Mmm, I was hoping you'd say that.
Feels electric already, doesn't it?
It really does, your smile is magnetic
Your choice…
Lives in the Chrome sidepanel — visible while you browse
Bring Your Own Everything
Scrya never locks you in. Swap out any layer — the brain, the voice, the characters.
Bring Your Own LLM
Point Victoria at any OpenAI-compatible endpoint. Ollama running locally, LiteLLM proxy, or any hosted model — if it speaks the API, it works.
Don't Like Grok's Voice?
Swap in any of 150+ Eleven Labs community voices — or use Chatterbox/Higgs Audio for fully local, uncensored TTS. Voice plays beat-by-beat as replies land.
Bring Your Own Characters
Import characters from any IMDB film — the story DAG seeds their personality, wants, fears, and arc. Or build your own persona with the character builder.
Replies That Feel Real
Victoria doesn't dump a wall of text. Long replies split into natural-sized beats — each one arrives with a typing indicator, then lands and plays its voice clip before the next begins.
- ✓ Typing dots between each beat — Just like iMessage — you see her thinking
- ✓ Beat-synced audio — Voice plays as each beat lands, not at the end
- ✓ Refresh choice buttons — Regenerate suggested replies any time
- ✓ End Turn actions — Victoria can autonomously act, inject a video prompt, or end the scene
- ✓ Jailbreak toggle — Full uncensored roleplay when you need it
You making me blush already.
What should we do with all this spark between us?
9 Languages, Same Character
Victoria follows your system language automatically. No manual toggle — she just switches.
What's Under the Hood
From the commit log — shipped, not planned
LiteLLM proxy + auto key provisioning
Scrya LLM gateway preset — one-click configure against api.scrya.com. Auto-provisions virtual LiteLLM keys. HERETIC NVFP4 quantization pipeline for local GPU inference.
Per-backend voice-dialect overrides
The LLM emits control syntax the active TTS backend understands. Paralinguistic markers ([laugh], [sigh]) are stripped before synthesis so they don't get read aloud.
Method-acting persona briefs
IMDB story DAG fields — wants, fears, secret, arc — are LLM-expanded into a rich system prompt. The character doesn't just know their name; they know their wound.
Video DAG cache from Victoria replies
Victoria's responses feed a video DAG cache — her emotional state and scene context are available to the video prompt pipeline for consistent image generation.
Four-provider action chain
End Turn uses a two-stage category → action selection pipeline with a 1,000+ action vocab, world boosted_keywords injection, and temperature=0.9 for diverse outputs.
Per-character system prompt editor
Full editor for Victoria's persona brief. Chatbots imported from IMDB are locked to prevent accidental edits to their canonical character definition.
Your companion is waiting.
Free Chrome extension. No account required to start. Bring your own LLM, voice, and characters — or use the defaults and go.
Install Free from Chrome Web Storev6.7.1 · Free · Chrome