{ "version": 2, "updatedAt": "2026-09-01", "plugins": [ { "id": "mlx-vlm", "kind": "asr", "name": "MLX-VLM", "status": "ready", "install": { "kind": "bundle", "id": "mlx-vlm" }, "notes": "External JSON-RPC provider for Gemma 4 and other mlx-vlm audio models. Install with `vox plugins install mlx-vlm`, then restart voxd." } ], "models": [ { "id": "parakeet:v3", "kind": "asr", "family": "parakeet-tdt", "name": "Parakeet TDT v3", "vendor": "nvidia", "runtime": "coreml", "status": "ready", "default": true, "languages": "25 European languages", "notes": "Default on-device dictation model. Multilingual FastConformer-TDT.", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": true }, "source": { "type": "huggingface", "repo": "FluidInference/parakeet-tdt-0.6b-v3-coreml" }, "parakeet": { "cacheDirectoryName": "parakeet-tdt-0.6b-v3", "jointFile": "JointDecisionv3.mlmodelc", "vocabularyFile": "parakeet_vocab.json", "blankId": 8192, "requiredFiles": [ "Preprocessor.mlmodelc", "Encoder.mlmodelc", "Decoder.mlmodelc", "JointDecisionv3.mlmodelc" ] } }, { "id": "parakeet:v2", "kind": "asr", "family": "parakeet-tdt", "name": "Parakeet TDT v2", "vendor": "nvidia", "runtime": "coreml", "status": "ready", "languages": "English", "notes": "English-only TDT with a tighter vocabulary and better English recall.", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": true }, "source": { "type": "huggingface", "repo": "FluidInference/parakeet-tdt-0.6b-v2-coreml" }, "parakeet": { "cacheDirectoryName": "parakeet-tdt-0.6b-v2", "jointFile": "JointDecision.mlmodelc", "vocabularyFile": "parakeet_vocab.json", "blankId": 1024, "requiredFiles": [ "Preprocessor.mlmodelc", "Encoder.mlmodelc", "Decoder.mlmodelc", "JointDecision.mlmodelc" ] } }, { "id": "apple:speech-transcriber", "kind": "asr", "family": "apple-speech", "name": "Apple SpeechTranscriber", "vendor": "apple", "runtime": "speech-analyzer", "status": "ready", "languages": "System-supported SpeechTranscriber locales", "notes": "Core on-device parity option. Apple manages the model assets and audio format negotiation; Vox currently returns finalized file transcription and word timings.", "platforms": ["macOS 26+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": true }, "source": { "type": "system" } }, { "id": "moonshine:medium-streaming", "kind": "asr", "family": "moonshine", "name": "Moonshine Medium Streaming", "vendor": "moonshine-ai", "runtime": "moonshine-swift", "status": "ready", "languages": "Configurable; English by default", "notes": "Native next-generation streaming architecture through MoonshineVoice. Vox currently exposes it through the finalized file-transcription contract; VOX_MOONSHINE_LANGUAGE selects the downloaded language model.", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "package", "repo": "moonshine-ai/moonshine-swift" } }, { "id": "mlx-community/parakeet-tdt-0.6b-v3", "kind": "asr", "family": "mlx-audio", "name": "Parakeet TDT v3 (MLX)", "vendor": "nvidia", "runtime": "mlx-audio", "status": "ready", "languages": "25 European languages", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/parakeet-tdt-0.6b-v3" } }, { "id": "mlx-community/parakeet-tdt-0.6b-v2", "kind": "asr", "family": "mlx-audio", "name": "Parakeet TDT v2 (MLX)", "vendor": "nvidia", "runtime": "mlx-audio", "status": "ready", "languages": "English", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/parakeet-tdt-0.6b-v2" } }, { "id": "mlx-community/whisper-large-v3-turbo-asr-fp16", "kind": "asr", "family": "mlx-audio", "name": "Whisper Large v3 Turbo", "vendor": "openai", "runtime": "mlx-audio", "status": "ready", "languages": "99+ languages", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/whisper-large-v3-turbo-asr-fp16" } }, { "id": "mlx-community/Qwen3-ASR-1.7B-8bit", "kind": "asr", "family": "mlx-audio", "name": "Qwen3-ASR 1.7B", "vendor": "qwen", "runtime": "mlx-audio", "status": "ready", "languages": "30 languages and 22 Chinese dialects", "notes": "Recommended Qwen3 ASR size on Apple Silicon. The current Vox mlx-audio bridge is file-based.", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/Qwen3-ASR-1.7B-8bit" } }, { "id": "mlx-community/Qwen3-ASR-0.6B-8bit", "kind": "asr", "family": "mlx-audio", "name": "Qwen3-ASR 0.6B", "vendor": "qwen", "runtime": "mlx-audio", "status": "ready", "languages": "30 languages and 22 Chinese dialects", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/Qwen3-ASR-0.6B-8bit" } }, { "id": "mlx-community/cohere-transcribe-03-2026-mlx-8bit", "kind": "asr", "family": "mlx-audio", "name": "Cohere Transcribe 03-2026", "vendor": "cohere", "runtime": "mlx-audio", "status": "ready", "languages": "14 languages", "notes": "Local MLX conversion of Cohere Transcribe. Offline rather than realtime; requires the mlx-audio ASR provider.", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/cohere-transcribe-03-2026-mlx-8bit" } }, { "id": "mlx-community/nemotron-3.5-asr-streaming-0.6b-8bit", "kind": "asr", "family": "mlx-audio", "name": "Nemotron 3.5 ASR Streaming 0.6B", "vendor": "nvidia", "runtime": "mlx-audio", "status": "ready", "languages": "35 languages", "notes": "Cache-aware streaming FastConformer-RNNT with configurable latency. The model is streaming-capable; the current Vox mlx-audio bridge is file-based.", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/nemotron-3.5-asr-streaming-0.6b-8bit" } }, { "id": "gpt-transcribe", "kind": "asr", "family": "openai-transcribe", "name": "GPT Transcribe", "vendor": "openai", "runtime": "remote", "status": "ready", "languages": "multilingual", "notes": "Current OpenAI file-transcription model.", "requires": ["OPENAI_API_KEY"], "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": false, "wordTimestamps": false } }, { "id": "gpt-4o-transcribe", "kind": "asr", "family": "openai-transcribe", "name": "GPT-4o Transcribe", "vendor": "openai", "runtime": "remote", "status": "ready", "languages": "multilingual", "requires": ["OPENAI_API_KEY"], "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": false, "wordTimestamps": false } }, { "id": "gpt-4o-mini-transcribe", "kind": "asr", "family": "openai-transcribe", "name": "GPT-4o Mini Transcribe", "vendor": "openai", "runtime": "remote", "status": "ready", "languages": "multilingual", "requires": ["OPENAI_API_KEY"], "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": false, "wordTimestamps": false } }, { "id": "whisper-1", "kind": "asr", "family": "openai-transcribe", "name": "Whisper-1", "vendor": "openai", "runtime": "remote", "status": "ready", "languages": "multilingual", "notes": "Legacy OpenAI Whisper API. The current Vox bridge uses finalized file transcription and does not yet request its timestamp or caption response formats.", "requires": ["OPENAI_API_KEY"], "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": false, "wordTimestamps": false } }, { "id": "gemma-4-e2b-it", "kind": "asr", "family": "mlx-vlm", "name": "Gemma 4 E2B", "vendor": "google", "runtime": "mlx-vlm", "status": "plugin", "plugin": "mlx-vlm", "languages": "multilingual", "notes": "Install with `vox plugins install mlx-vlm`. The plugin speaks the Vox provider protocol and runs Gemma through mlx-vlm.", "platforms": ["macOS 14+"], "architectures": ["arm64"], "capabilities": { "fileTranscription": true, "liveTranscription": false, "onDevice": true, "wordTimestamps": false }, "source": { "type": "huggingface", "repo": "mlx-community/gemma-4-E2B-it-4bit" } } ] }