Galileo Avatar

Live AI avatars built around your knowledge, your guardrails, and your face.

Two tiers. Preset marketing avatars in a week. Conversational avatars with custom knowledge in four to six. Sydney-built on Anthropic Claude, HeyGen, and ElevenLabs. End-to-end, fully auditable.

$15K–$25K

Preset, 1 week

$25K–$75K

Conversational, 4–6 weeks

End-to-end

Not a wrapper

100%

Audit trail

See case studies

Two tiers, one custom build

Pick the avatar your use case needs.

Both tiers ship as a custom Claude Code application built around a multi-vendor stack. Anthropic Claude, HeyGen Streaming, ElevenLabs ConvAI, Supabase audit, your brand. The difference between tiers is whether the avatar follows a script or talks back.

Preset Marketing Avatar

$15K–$25K· 1 week to live

A scripted, on-brand video avatar that delivers your message at scale, in any language, on any channel.

Best for

Outbound sales videos
Training and onboarding modules
Product walk-throughs
Multilingual launches
Executive comms at scale

Includes

Branded avatar from your imagery
Scripted persona, brand-voice locked
Multilingual delivery
Video export pipeline
Brand-safety guardrails

Stack

Anthropic ClaudeHeyGen videoElevenLabs voiceCustom Next.js appBrand pipeline

Four things off-the-shelf avatars cannot do.

Custom knowledge, not generic Q&A

We ingest your firm's documents, policies, and process maps. The avatar answers from your knowledge, not the public internet.

Guardrails you write

You decide what the avatar can say, what it must refuse, and what it escalates to a human. No surprises, no off-script answers.

Your face, your voice, your brand

Avatars are built from your team's imagery and voice samples. The persona feels like part of your firm, not a stock template.

Audit trail built in

Every session is transcribed, summarised, and timestamped. Replay any conversation, defensible for compliance and HR contexts.

What we build

Galileo-built avatar patterns ready for your sector.

Four in production today, one pattern ready to build. Every one a custom Claude Code application: your brand, your knowledge, your guardrails, your audit trail.

Engineering · Graduate assessment

In production

Cut graduate-assessment interviewer hours by 60% with a live avatar pipeline.

A 7,000-person Australian engineering consultancy runs its graduate assessment centre through a Galileo-built avatar. Voice-driven structured interview, real-time scoring, multi-device deployment, branded end-to-end. Now in continuous production.

200+

candidates assessed

HeyGen StreamingElevenLabs ConvAIAnthropic ClaudeCustom Next.js appSupabase audit

Galileo Logic · AI Business Analyst

In production

Galileo's own AI Business Analyst runs the discovery interview for every $50K engagement.

We use the same stack we sell. A custom-trained ElevenLabs persona conducts the AI readiness interview that opens every Galileo Discovery and Cowork rollout. Trained per client on their tech stack, pain points, and target workflows. Outputs a costed team recommendation, ROI preview, and validation playbook the client keeps.

Every engagement

trained per client

Anthropic Claude (Opus + Sonnet)ElevenLabs ConvAIPer-client knowledge ingestionCustom Next.js appBusiness case PDF

Same stack runs in production at scale: ElevenLabs for customer support, HeyGen for training video, Anthropic Claude for the AI applications underneath. We package it for Australian firms.

Where this fits

Avatar patterns mapped to your industry.

Concrete avatar applications for each of the ten niches we recruit and build for.

How it runs

Scope to live in one to six weeks.

Same four steps for both tiers. Timings stretch with knowledge depth, not with process.

Scope

30 min

Discovery call to define the avatar's job. We capture the use case, the audience, the success metric, and the source material. Choose human-like or stylised avatar. Choose persona, voice, and language coverage. Identify the knowledge sources we ingest for RAG, the guardrails the avatar must obey, and the systems we integrate (ATS, SSO, CRM). You leave with a one-page scoping doc you keep regardless of whether you proceed.

Build

1 week (Preset) · 3–4 weeks (Conversational)

A custom Claude Code application built around the avatar engine. We ingest your documents, build the RAG index, define the system prompt and guardrails, clone the voice, configure the HeyGen avatar with your branding, and wire the Next.js front-end. Integrations: HeyGen Streaming SDK, ElevenLabs ConvAI, Anthropic Claude (Opus for reasoning, Sonnet for routing), Supabase for the audit pipeline, plus any client systems (SSO, ATS, CRM). We tune response latency, test against real-world prompts, and harden the guardrails iteratively. Not a wrapper; a fit-for-purpose application.

Pilot

1 week

The avatar runs against a controlled cohort or audience while we watch. Real sessions, real transcripts, real edge cases. We review the scoring rubric, calibrate prompt strength against actual usage, fix guardrail gaps the cohort surfaces, and pressure-test latency under load. You see the full audit log and the session-by-session summaries. Adjustments land before production opens.

Production

Ongoing

Multi-device deployment with uptime monitoring, error alerting, and the audit pipeline writing to your retention store. Voice and inference run at $0.50 to $2.00 per conversation depending on length and language. We hand over runbooks, on-call coverage for the first month, and a quarterly review cadence for prompt drift, knowledge refresh, and guardrail evolution. The avatar improves over its lifetime, not on day one.

Live avatars in 4–6 weeks

Pilot an avatar.

Tell us your use case. 30-minute scoping call, fixed-price quote, live in one to six weeks. End-to-end build, your IP.

See pricing