1. Smart Import & Automated Preprocessing

Effortlessly handle individual files, batches, or deeply nested folder trees with background scanning and zero-friction setup. When media is added, WaveCap automatically standardizes audio to a high-fidelity 44.1 kHz 16-bit mono WAV format to maximize pipeline speed and maintain predictable system memory usage. During import, bundled Core ML models—Silero VAD v6.2.1 and ECAPA-TDNN VoxLingua107—run entirely on the CPU to evaluate chronological speech blocks and identify spoken languages without taxing your GPU or Neural Engine. This intelligent pre-analysis ensures your media is perfectly formatted and tagged before full processing begins, all while keeping your Mac awake and responsive.

2. Advanced Vocal Isolation & Audio Cleanup

Transform noisy, echo-filled, or poorly recorded media into studio-grade audio prior to transcription. WaveCap leverages state-of-the-art vocal separation and speech enhancement engines, including DeepFilterNet (v2/v3) and Demucs/MDX architectures, to strip away room reverb, wind, traffic, and background music. The system automatically evaluates background noise levels upon import to determine if cleanup is required, while giving you full manual override to force or disable processing per file. By isolating clean vocal stems and preserving background audio separately, WaveCap drastically boosts speech-to-text accuracy and retains clean ambient audio tracks for future mixing.

3. Precision Speech-to-Text (STT)

Convert spoken dialogue into word-accurate, timestamped text using over 20 state-of-the-art open-source speech recognition models. WaveCap gives you total freedom to select the exact model family that fits your project's balance of speed, memory usage, and language complexity—including Whisper (Tiny to Large v3 Turbo), Parakeet, Qwen3 ASR, Cohere Transcribe, SenseVoice, Voxtral, and Nemotron ASR. Native Apple Silicon and MLX runtime optimizations deliver up to 4x faster local inference, allowing you to transcribe long interviews, fast-paced podcasts, and dense lectures locally with pinpoint accuracy, zero cloud token costs, and total data privacy.

4. Local LLM Text Polishing & Translation

Refine raw transcriptions, correct grammatical errors, and translate content into over 50 global languages without sending a single byte to external APIs. Powered by local Large Language Models such as Qwen 3.5, Gemma 4, Llama 3.1, and Nemotron 3 Nano, WaveCap intelligently cleans up messy transcripts by fixing broken sentences, removing filler words, and correcting specialized domain terminology. You can seamlessly translate transcripts into full subtitle files while maintaining semantic context and formatting, ensuring high-quality multilingual localization with zero recurring subscription fees.

5. Speaker Diarization

Automatically distinguish, separate, and label individual voices across complex multi-speaker recordings. Driven by the Sortformer Streaming 4SPK diarization engine, WaveCap analyzes vocal characteristics throughout the audio timeline to identify speaker transitions and assign distinct labels for up to four unique speakers. Whether you are processing panel discussions, casual podcast conversations, or formal meeting recordings, speaker labels are seamlessly mapped to transcript segments and can be embedded directly into exported SRT/VTT subtitle files, plain text documents, CSVs, or structured JSON data.

6. Timeline-Aligned AI Dubbing (TTS)

Turn translated subtitles into realistic, fully localized voice tracks that stay locked to your original video timing. Utilizing local Text-to-Speech models like Qwen3 TTS, OmniVoice, and Breeze TTS, WaveCap synthesizes natural-sounding dialogue mapped precisely to each subtitle segment's timestamps, using original audio or diarized speaker clips as voice references. When vocal separation is enabled, WaveCap can recover and re-mix the original background audio and music behind the new voiceover, allowing you to export standalone audio files or a fully dubbed MOV video that replaces original speech while preserving video quality and ambient soundscapes.