Tech • IA • Crypto
Tech news for 22/05/2026 — Live on Renaud Dékode
4/10TL;DR
A French tech community is launching Cast, a real-time voice AI tool, expanding an ecosystem that emphasizes accessible, local-first usage and collaborative development.
KEY POINTS
Launch of Cast, a real-time voice AI tool
A new tool called Cast enables real-time voice interactions with AI models, focusing on conversational immediacy rather than traditional text-based exchanges. It leverages real-time inference instead of standard speech-to-text and text-to-speech pipelines, aiming to create more natural dialogue and responsive “personas.”
Built on community-driven experimentation
Cast originated from collaborative workshops where participants iteratively built and refined prototypes. This approach mirrors the earlier development of Kiro, another tool in the ecosystem, highlighting a model of co-creation where user feedback directly shapes product evolution.
Kiro platform expands multi-AI access
The existing Kiro platform aggregates multiple AI models into a single interface, allowing users to switch providers without subscriptions. It supports local, sovereign usage, direct API connections, and pay-as-you-go access, positioning itself as an alternative to centralized AI services.
Focus on accessibility and low barriers
Both tools are distributed with minimal friction: no installation, optional costs, and compatibility with free or local models. The initiative emphasizes democratizing AI usage rather than premium, high-cost training ecosystems that can reach thousands of euros.
Real-time AI seen as underutilized
The developers argue that real-time AI remains underestimated compared to conventional pipelines. By enabling dynamic conversations, Cast aims to unlock new use cases such as training simulations, customer service practice, and interactive content creation.
Technical limitations shape model support
Not all AI providers are currently compatible with real-time voice features. Models lacking native real-time capabilities, such as some offerings from Anthropic or Mistral, are excluded for now, while others like Google’s models are favored for cost efficiency and performance.
Local data handling and security considerations
The tools prioritize client-side storage, keeping API keys and data within the user’s browser environment. While this improves privacy, users are still advised to manage keys carefully and consider security practices depending on their setup.
Rapid iteration through user feedback
Cast is released in beta, with planned improvements driven by user suggestions. Previous experience with Kiro showed that community input can quickly enhance usability and feature depth, accelerating development cycles.
Broader ecosystem under development
Additional tools and platforms are reportedly in progress, suggesting the emergence of a broader AI toolkit ecosystem built around experimentation, education, and practical deployment rather than standalone products.
CONCLUSION
The rollout of Cast reflects a growing shift toward real-time, user-driven AI tools that prioritize accessibility, privacy, and rapid iteration over centralized, high-cost solutions.