Versions:
Hablará, developed by Fidpa, is an AI-powered conversation analysis application currently at version 1.7.7, with five versions released to date. The software records conversations, transcribes them locally, and analyses emotions and communication styles without requiring cloud connectivity or a subscription. Its core analytical capabilities include dual-track emotion detection from voice and text, combining audio analysis with LLM processing, and drawing on established psychological frameworks such as Plutchik's Wheel of Emotions and Russell's Circumplex Model. Additional analyses cover tone across five dimensions and topic classification into seven categories, while advanced options extend to logical fallacies, nonviolent communication, cognitive distortions, the four-sides model, transactional analysis, appraisal theory, regulatory focus, and conversation analysis. A distinctive reflection feature offers Bible Impulse, presenting one to three thematically matching Bible verses for a transcript, alongside AI-generated prayers tailored to content and emotion. The application also includes an AI chat component with automatic conversation summaries, contextual discussions about recordings, and guidance about the app itself. Export functionality supports PDF documents containing transcriptions and analyses, chat exports in five formats (Markdown, TXT, PDF, HTML, and DOCX), and recording downloads as WAV files. Hablará operates fully offline, using whisper.cpp for local transcription and Ollama as the default free local AI provider, though users may optionally configure cloud-based LLMs from OpenAI, Anthropic Claude, or Mistral, the latter offering EU data residency. Privacy and GDPR compliance are emphasized through API key encryption in the Windows Credential Manager, local storage of recordings, configurable automatic deletion, and opt-in anonymous crash reporting via EU servers. The software suits users in counselling, coaching, pastoral care, communication training, and personal reflection contexts who require confidential, locally processed analysis of spoken interactions.
Tags: