Build a Fully On-Device Voice ChatGPT for iOS with Apple's Free Models
amos_gyamfi · x · 2026-09-03
Developer Amos Gyamfi shares a fully local, on-device iOS AI assistant stack: text generation via Apple Intelligence's Foundation Models, speech-to-text via the SpeechAnalyzer and SpeechTranscriber models, and voice output via the 82M-parameter open-source Kokoro TTS.
All three stages — speech recognition, text generation, and speech synthesis — run offline on-device with no cloud API required. Links to try it out are included.
Related event: Build a Local ChatGPT on iOS with Apple's Free Models(2 posts)→
More from coding & agent
- Free n8n workflow auto-pulls weekly ad reports across all platforms via Databox MCP — aftahi_ai · 2026-09-03
- 用 TRL+OpenEnv 开源复现「写代码画水彩」模型全流程 — huggingface · 2026-09-03
- Developers Praise Claude Design: Great Results with a Design System — KlausCodes · 2026-09-03
- Anthropic engineer on building self-improving AI systems: loops and graphs explained — goyalshaliniuk · 2026-09-03
- Wes Roth Builds Four Full AI Games on Claude Fable 5.1's Low-Effort Setting — Wes Roth · 2026-09-03
- Dev swaps in Cursor and Grok 4.6 subagents when hitting Claude Code limits to save cost — rudrank · 2026-09-03