LiveTranscriber: Open-Source iOS App Runs Whisper, Qwen3-ASR & LLMs Fully Offline
marshmallow_ki · reddit · 2026-08-06
A developer has introduced LiveTranscriber, an open-source iOS app that manages to run modern speech and language models entirely on-device and offline on an iPhone, pushing past mere tech demos into practical utility.
Key Features & Supported Models:
- Offline Transcription: Supports Whisper, Qwen3-ASR, and NVIDIA Nemotron Streaming for low-latency live transcription.
- Speaker Awareness: Integrates MOSS Multi-Speaker for speaker-aware transcription.
- Local Analysis: Utilizes Qwen3 to generate summaries, extract key points, and analyze transcripts directly on the device.
- Ecosystem: Includes Apple Watch recording with automatic sync, real-time translation, and searchable history.
The author noted that the primary engineering hurdles weren't just running the models, but optimizing mobile memory management, streaming latency, inference backend switching, and battery consumption.
Related event: Open-source iOS App LiveTranscriber Runs Multiple Models Offline(2 posts)→
More from Infra
- Discussion: Running llama-server Inference Across Machines via RPC Clustering — _TheWolfOfWalmart_ · 2026-08-06
- Modal Rebuilds Sandbox Platform to Create 1M Concurrent Containers in Under a Minute — dscape · 2026-08-06
- Nebius Tops Endpoint Accuracy for GLM-5.2, Hits ~300 Tokens/s Output — songhan_mit · 2026-08-06
- Gavin Baker on AI Compute: SRAM Accelerators Offer Unbeatable ROI, Disaggregation is Key — IanAndrewsDC · 2026-08-06
- Using M4 Max MacBook as an Always-On LLM Server for Mobile Devices — michaelthatsit · 2026-08-06
- ExANS: Open-Source Lossless KV Cache Compression Hits 622 GB/s on H100 — arnav__1 · 2026-08-06