Offline • Low‑Latency • Real‑Time • Faster‑Whisper Engine
transcribe-assistant is an offline, low-latency (~250 ms) real-time speech-to-text engine built for automation workflows, assistants, accessibility tools and privacy‑sensitive environments.
It uses local Faster‑Whisper inference, provides zero cloud dependencies, supports session recording, modular backend architecture, and includes a deterministic PRO version for compliance‑ready use cases.
transcribe-assistant is a fast, offline, real-time transcription engine built for:
- automation workflows
- assistants and agents
- accessibility
- offline environments
- low-latency speech-to-text
This is the public version (non-deterministic).
The deterministic PRO version is part of the Krunixbase ecosystem.
- ~250 ms latency
- offline inference
- Faster-Whisper backend
- streaming transcription
- session recording
- export to TXT / JSON
- CLI + Python API
See full architecture here:
ARCHITECTURE.md
pip install -r requirements.txt
python main.py
See API documentation:
API.md
See full roadmap:
ROADMAP.md
See pricing model:
PRICING.md
MIT License.