Soul Machines

Editorial notes

Soul Machines at a glance

Soul Machines packages conversational avatar work as interactive AI assistants, Digital People, Studio, and Digital Workforce. Studio is the creation and deployment environment; its pricing page lists templates, gestures, behavior configuration, voice and language configuration, conversational-AI integration, reporting, and live deployment. Digital Workforce adds a workflow-focused product, with public material mentioning integrations such as Zapier and ServiceNow. A buyer should evaluate the whole agent: purpose, information source, interaction design, workflow permissions, and post-session reporting, rather than treating a digital face as a video asset.

The company offers no-code and developer routes. Its developer site documents a Web SDK for frontend integration and a Skills API for web services that respond to conversational input. A team can start from a template and later connect a customer-owned service, but must decide which system is authoritative. A skill that takes action needs narrow credentials, auditable inputs and outputs, explicit failure behavior, and a human fallback. A welcoming avatar does not reduce the security review needed for an account lookup, HR process, or support automation.

Read the tier boundaries

Studio’s free tier supports exploration of assistant creation, templates, gestures, behavior, and conversational integration, but its pricing page lists zero interactive-conversation minutes. It should not be called a free live-agent allowance. The same official page is internally inconsistent: plan cards advertise a seven-day trial, while its subscription FAQ says 14 days. Confirm the current trial length directly with Soul Machines before relying on either statement. Basic lists 40 interactive minutes and one live assistant, Plus lists 350 minutes and up to three assistants, and Pro lists 10,000 minutes and six assistants; enterprise terms go through sales. Confirm trial conditions and plan access in the portal before relying on an allowance.

Interactive minutes and live deployment support native real time, but they do not imply a latency guarantee. Public material uses real-time engagement language, while reviewed sources do not publish a numerical universal response-time SLA. Measure the full turn: recognition, reasoning, connected skills, speech, animation, and delivery. Natural gestures can make a delay feel different, but visual quality is not a performance commitment.

Design for the role

Start with one bounded role: explain a benefit, collect issue details, or practice a customer conversation. Define what the assistant may say, do, retrieve, and hand off. Test denied permission, stale CRM data, unavailable integration, duplicate requests, and changed intent. For training, test whether feedback is grounded in actual learner statements. Keep a clear disclosure that the participant is talking with an automated agent.

Pricing also lists downloadable video alongside interactive conversations. Budget and review those separately. A clip can be approved before publication; a live assistant needs monitoring, privacy-safe observability, and incident response. For a likeness, obtain consent and a retirement path.

Decision guidance

Choose Soul Machines when a team wants a creation environment, live deployment, and a route to custom frontend or service integrations. Tavus, Simli, and Anam are more direct API-led alternatives. UneeQ is a closer comparison for immersive role-play. Compare candidates with identical knowledge, escalation rules, and task metrics.

A launch owner should review every template change and connected workflow before publication. Measure completion, escalation, and correction rates over multiple sessions, then compare those outcomes with a non-avatar process. The right investment case is better service or practice quality, not the novelty of an animated interface.