Right Call is an AI-powered communication platform built around specialized voice models and intelligent agent orchestration. It is designed to connect natural conversation with context, routing, and business workflows—without blurring research goals, performance targets, and verified production claims.
The Delulu model family
One voice intelligence ecosystem. Three distinct models.
From premium multilingual intelligence to efficient on-device conversations, the Delulu family explores voice AI across different deployment environments.
01
Premium Voice AI Model
Delulu MAX
Availability to be confirmed
Our premium voice intelligence offering, designed for high-performance conversational experiences and multilingual communication.
DeploymentManaged / cloud-supported
PerformanceTarget conversational latency: under 350 ms.
Advanced AI-agent orchestration
Premium real-time conversational voice experience
Indian-language and dialect-aware interactions
Advanced context management and intelligent task routing
Natural conversational turn-taking
Integration with Right Call services
02
Local Voice AI Model
Delulu Lite
Availability to be confirmed
A lightweight voice intelligence model designed to run locally through the Right Call Desktop application.
DeploymentLocal desktop
PerformancePerformance target: under 390 ms. Test evidence pending.
Local execution through Right Call Desktop
Low-latency voice interaction by design
Malayalam, five additional Indian languages, and Indian English
Offline processing after installation, subject to deployment configuration
Hardware-aware optimization
Local-first privacy architecture
Local AI. Local processing. Intelligent conversations.
Validation record
Verified test hardware
Pending validation
Runtime
Pending validation
Quantization
Pending validation
Latency methodology
Pending validation
03
Research Model
Delulu Nano
Under Research
An ultra-compact research initiative exploring practical conversational AI capabilities on consumer smartphones.
DeploymentTargeted on-device
PerformanceTechnical specifications: to be announced.
Smartphone-local inference research
Malayalam and English research targets
Lightweight model architecture
Energy-efficient processing exploration
Reduced memory requirements
Practical on-device conversational interaction
Research preview — not an available commercial release.
Model comparison
Choose the deployment path, not just the model size.
Targets and research goals are separated from independently verified performance claims.
Comparison of Delulu MAX, Delulu Lite, and Delulu Nano
Capability
Delulu MAX
Delulu Lite
Delulu Nano
Positioning
Premium
Local Desktop
Mobile Research
Deployment
Managed / cloud-supported
Local Desktop
Targeted on-device
Malayalam
Subject to validation
Supported
Research target
Indian languages
Multilingual
Malayalam + 5 others
Malayalam and English
Indian English
Supported
Supported
English research target
Latency
Under 350 ms target
Under 390 ms target; evidence pending
TBA
Platform
Right Call ecosystem
Right Call Desktop
Smartphones — research target
Status
Confirm availability
Confirm availability
Under Research
Premium Voice AI ModelDelulu MAX
Positioning
Premium
Deployment
Managed / cloud-supported
Malayalam
Subject to validation
Indian languages
Multilingual
Indian English
Supported
Latency
Under 350 ms target
Platform
Right Call ecosystem
Status
Confirm availability
Performance values remain targets until the team publishes verified measurements, test conditions, hardware or network environment, and a clear latency definition.
Agent architecture
Beyond voice. An intelligent orchestration layer.
Right Call coordinates speech processing, contextual intelligence, task routing, and response generation through a modular agent architecture.
Conceptual demonstration
Selected pathDelulu MAX
DeploymentManaged / cloud-supported
Language scopePremium multilingual model family; definitive language list pending validation.
PerformanceTarget conversational latency: under 350 ms.
Managed orchestration path for multilingual business communication workflows.
Conversation experience
Designed for natural conversations.
A responsive visual study of speech rhythm, turn-taking, and streaming response behavior. This is an animation preview, not a live conversation or measured system output.
Animation Preview
Preview runningNo microphone access
Performance methodology
Engineered for responsive conversations.
Perceived voice latency is a pipeline, not a single number. Transparent reporting separates first-audio latency, model inference, and full-response completion.
01
Capture + buffering
Microphone capture, audio framing, and buffer strategy shape how quickly useful speech reaches the system.
02
Speech processing
The processing path may be modular or integrated depending on the selected model and deployment.
03
Reasoning + response
Context, routing, and response generation contribute separately to the time a conversation feels like it takes.
04
Synthesis + streaming
First audio can stream before a full response is complete; these measurements should not be conflated.
05
Network transport
Cloud-supported deployments also depend on connectivity, distance, congestion, and telephony infrastructure.
