Modellen en licenties
Vocatim gebruikt modellen en bibliotheken van andere makers, onder hun licenties. De modellen in Vocatim zijn aangepast ten opzichte van de originelen: omgezet naar Core ML of gekwantiseerd naar GGUF.
Modellen
Parakeet TDT 0.6B v3
- Origineel: huggingface.co/nvidia/parakeet-tdt-0.6b-v3
- Omzetting naar Core ML: huggingface.co/FluidInference/parakeet-tdt-0.6b-v3-coreml
Nemotron 3.5 ASR Streaming 0.6B
- Origineel: huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b
- Omzetting naar Core ML: huggingface.co/FluidInference/Nemotron-3.5-ASR-Streaming-Multilingual-0.6b-CoreML
Sortformer v2.1
Kennisgeving: Licensed by NVIDIA Corporation under the NVIDIA Open Model License
- Origineel: huggingface.co/nvidia/diar_streaming_sortformer_4spk-v2.1
- Omzetting naar Core ML: huggingface.co/FluidInference/diar-streaming-sortformer-coreml
- Bron: NVIDIA Trustworthy AI terms
pyannote speaker-diarization-community-1
Bevat:
- pyannote speaker-diarization-community-1: CC BY 4.0
- PLDA parameters, BUT Speech@FIT: CC BY 4.0
- WeSpeaker ResNet34-LM (WeSpeaker, pyannote): CC BY 4.0
- pyannote segmentation-3.0: MIT, Copyright (c) 2023 CNRS
- Origineel: huggingface.co/pyannote/speaker-diarization-community-1
- Origineel: huggingface.co/pyannote/segmentation-3.0
- Origineel: huggingface.co/pyannote/wespeaker-voxceleb-resnet34-LM
- Omzetting naar Core ML: huggingface.co/FluidInference/speaker-diarization-coreml (NOTICE.md)
Whisper base, large-v3, large-v3-turbo
Bevat:
- Whisper base, Whisper large-v3: Apache License 2.0
- Whisper large-v3-turbo: MIT
- Core ML conversion, Argmax: MIT
- Origineel: huggingface.co/openai/whisper-base
- Origineel: huggingface.co/openai/whisper-large-v3
- Origineel: huggingface.co/openai/whisper-large-v3-turbo
- Omzetting naar Core ML: huggingface.co/argmaxinc/whisperkit-coreml
Qwen3.5 2B, 4B, 9B
- Origineel: huggingface.co/Qwen/Qwen3.5-2B
- Origineel: huggingface.co/Qwen/Qwen3.5-4B
- Origineel: huggingface.co/Qwen/Qwen3.5-9B
- GGUF-kwantisering: huggingface.co/unsloth/Qwen3.5-2B-GGUF
- GGUF-kwantisering: huggingface.co/unsloth/Qwen3.5-4B-GGUF
- GGUF-kwantisering: huggingface.co/unsloth/Qwen3.5-9B-GGUF
Qwen3 Embedding 0.6B
- Origineel: huggingface.co/Qwen/Qwen3-Embedding-0.6B
- GGUF-kwantisering: huggingface.co/DefaultDF/Qwen3-Embedding-0.6B-GGUF
Bibliotheken
FluidAudio 0.15.6
Bevat:
- fastcluster: BSD 2-Clause, © 2011 Daniel Müllner; changes from 1.1.24 on © Google Inc.
- VBx: Apache License 2.0, © 2021–2024 BUT Speech@FIT
text-processing-rs 0.3.0
Bevat:
- NVIDIA NeMo Text Processing: Apache License 2.0, © NVIDIA CORPORATION & AFFILIATES
- rustfst 1.3.1, flate2 1.1.9, anyhow 1.0.103, lazy_static 1.5.0: MIT or Apache License 2.0
- miniz_oxide 0.8.9: MIT, Zlib or Apache License 2.0
- nom 7.1.3: MIT
Kennisgeving: text-processing-rs, Copyright 2026 FluidInference
WhisperKit
Bevat:
- WhisperKit: MIT, © 2024 argmax, inc.
- swift-transformers (Hugging Face), modified by Argmax: Apache License 2.0, © 2022 Hugging Face SAS
llama.rn 0.12.9, llama.cpp / ggml
React Native, React, Expo, React Navigation, react-native-gesture-handler, react-native-screens, react-native-safe-area-context
Unicode Character Database
- Bron: unicode.org/license.txt