Conversational Voice App for Scalable Emotional Support

Overview

Soulace provides emotional de-escalation tools to youth and adults navigating anxiety. Their pilot started as a text-based chatbot to test core conversation flows. To broaden accessibility and support real-time somatic regulation, Soulace needed to transition to a scalable voice system.

By testing simple prototypes first, we helped the client avoid building expensive features too early. 26bitz engineered a secure cross-platform audio application distributed across Telegram Mini Apps, mobile apps, and a PWA. This turned early emotional support concepts into a functional voice system that isolates user data to comply with privacy regulations while maintaining a natural conversation flow.

Challenges of real-time processing

While text-based LLM interfaces are simple to deploy, transforming them into voice systems introduces technical hurdles around data privacy and conversation pacing.

We helped Soulace resolve critical bottlenecks:

• Pacing & Pauses: Breathing exercises require step-by-step progressions and deliberate pauses. Generative models stream text probabilistically and cannot hold or regulate speech delivery without external syntax control.
• Audio Latency: The initial system caused sluggish interactions resulting in conversational delays.

Conversational AI voice platform

26bitz architected the safe, efficient voice infrastructure, focusing on core components:

• Asynchronous Voice streaming: We switched to bidirectional WebSockets and integrated the OpenAI Realtime API to handle continuous audio data.
• Fragment-based Pacing: We implemented a data map containing 23 technique clusters and 96 anger-management exercises, structuring content into isolated fragments to programmatically control speech rates and enforce breathing pauses.
• Compliant infrastructure: We added age-verification checks, native account-purging mechanisms, and an OpenAI Moderation API workflow that redirects high-risk inputs to the 988 Crisis Line.
• Admin management: We built a custom back-office portal allowing staff to update prompts and rearrange techniques in real time using dynamic sorting tools.

Results

The Soulace platform successfully moved from a text-based prototype to a live voice ecosystem:

• 3x faster response: Compressed response delays from 12 seconds down to 4 seconds, delivering natural verbal speeds.
• 100% code reusability: A single repository maintains the web dashboard, Telegram Mini App, and mobile apps.
• Data privacy: A zero-trace data pipeline, automated data deletion, and recurring 7-day safety disclosures prepared the app for local regulatory approval.

Testimonial

"What stood out most about 26bitz was the rare combination of technical innovation and genuine compassion that drives their work. Finding a technology partner who not only excels in software development but also demonstrates a deep commitment to social impact and mental health advocacy is truly uncommon. The collaboration with 26bitz was seamless from start to finish. Communication, delivery, and professionalism consistently exceeded expectations, and all project goals were achieved on time and with exceptional quality."
— Joel Finch, President & Chair of the Board of Directors at Soulace Project

What's Next

With the core voice architecture and runtime established, the application is undergoing local regulatory checks. Once approved, the team will finalize remaining payment features and launch on official app stores.