A voice-based training tool for psychotherapy programs—built on the premise that human relationship is irreplaceable, and that preparation can make clinicians more present when it matters.
For training programs in psychodynamic, relational, humanistic, experiential, and integrative approaches.
Join the Research PartnershipThe consequence is familiar to every supervisor: trainees spend their first months managing their own anxiety rather than attending to the patient.
The patient—a real person who came seeking help—becomes the training ground for the trainee's developmental needs.
The Anteroom exists to reduce this burden.
Not to simulate therapy. Not to replace supervision. Not to train technique. To reduce the shock of first contact so clinicians can be more present when it matters.
Until recently, voice AI forced a choice: natural conversation or customizable content. Full-duplex speech models dissolve this constraint—and create possibilities for clinical training that didn't exist twelve months ago.
Half-duplex: strict turn-taking. The system listens, then speaks, then listens. Interruption breaks it. Silence is just waiting.
Simultaneous listening and speaking. The model attends to your voice while producing its own. Conversation flows like conversation.
When you interrupt the simulated patient, they respond—trailing off, pushing through, going silent. Your impulse to interrupt, and what happens next, becomes material for reflection.
In full-duplex interaction, a pause is not the system waiting. It's a patient not speaking. The phenomenological pressure to fill silence—the pull toward premature interpretation—can be experienced, not just described.
The pace at which a patient approaches and retreats from vulnerable material, the acceleration before dissociation, the slowing that accompanies genuine contact—temporal patterns that are invisible in text and impossible in half-duplex.
No re-reading. No pause button. The trainee must track, hold, and respond in the moment—the cognitive demand of actual clinical work, available for the first time in simulation.
Reading case descriptions is categorically different from hearing them. The information that matters for relational therapies—tone, rhythm, hesitation, affective texture—is carried in voice. Text cannot represent it. Trainees need practice perceiving it.
The controlled minimization of avoidant attachment. The pressured urgency of anxious relating. The flatness of shutdown. The fragmentation near dissociation. Heard, not described.
Intellectualization sounds a particular way—measured, affect-absent. Somatization has its texture. Defense is audible before it's nameable. Trainees develop ears for patterns.
Even knowing the patient is simulated, the body responds—with tension, protectiveness, withdrawal, boredom. These responses are data about the trainee's relational patterns.
Rather than promising a complete system, we're developing in tiers—validating each level before committing to the next. Full-duplex conversation is the aspiration; we build toward it through stages that are useful on their own.
Curated clinical vignettes with structured reflection exercises and supervisory discussion guides. High-quality voice synthesis, clinically validated presentations. Addresses perceptual exposure and vocabulary development. No interaction—but buildable and validatable now.
Interactive vignettes with choice points. The trainee listens, then selects or speaks an intervention; the patient's response branches accordingly. Active practice with clinical decisions, consequences visible. Proven interaction model, achievable with current technology.
Free-form spoken conversation with clinically coherent simulated patients. Natural interruption, silence, and rhythm. The full phenomenology of clinical encounter—minus the irreducible fact of another subject. This tier requires research: can current models maintain characterological consistency across extended interaction? We intend to find out.
This is not a product launch. It's an invitation to collaborative inquiry with training programs that share our premises about what matters in clinical education.
Partner programs help shape what gets built, participate in validation studies, and contribute clinical expertise that no technology team possesses alone. In return: early access, influence over development direction, and co-authorship on research publications.
We're particularly interested in: Psychoanalytic institutes. Doctoral programs in clinical and counseling psychology. Postdoctoral fellowships. Advanced psychotherapy training programs. Programs with strong supervisory cultures and interest in educational innovation.
Do experienced clinicians recognize these presentations as clinically coherent? This question is answered before development commits resources.
Deployment with incoming trainee cohorts. Mixed-methods evaluation of anxiety reduction, perceptual development, and supervisory engagement.
Decision gates at each phase. If the evidence says stop, we stop. Informative failure is still contribution to the field.
You pass through it on the way to somewhere more important. It doesn't need to be grand. It needs to be useful.
Begin a Conversation