Speech-to-Text

AI Models

Transcribe audio locally with zero API costs

• Fully offline after model download
• SRT, VTT, JSON, and text output
• Translate any language to English

What You Can Do

• Local transcription - Convert speech to text completely offline, no API key required
• Multiple model sizes - tiny (fastest) → base → small → medium → large (most accurate)
• Output formats - Plain text, SRT subtitles, VTT captions, or JSON with timestamps
• Translation mode - Translate any language audio directly to English text
• Wide format support - WAV, MP3, M4A, FLAC, OGG, and more
• Auto model caching - Downloads models on first use, fully offline after that

Try Asking

• "Transcribe this podcast.mp3 using the medium model"
• "Convert this interview to SRT subtitles"
• "Transcribe my voice memo and translate it to English"
• "Generate VTT captions for this video's audio track"
• "Use the large model for this important lecture recording"
• "Get JSON output with word-level timestamps"

Pro Tips

• tiny = fast but rough, small = good balance, medium = professional quality, large = maximum accuracy
• First run downloads the model (40MB–3GB depending on size), then fully offline
• SRT/VTT formats include timestamps for subtitle syncing
• Translation mode outputs English regardless of input language
• JSON output includes segment-level and word-level timing data
• Works completely offline after initial model download - great for privacy

Plans

More AI Models Skills

• Oracle - Bundle files and prompts to query any AI model
• Gemini - Quick one-shot queries to Google Gemini models
• Model Usage - Track AI model costs and token usage over time
• Whisper API - Fast cloud transcription via OpenAI Whisper API
• Things 3 - Full task management for Things 3 on Mac
• Discord - Send messages, polls, and reactions in Discord

Explore: Integrations · Glossary · Solutions · Use cases · Examples · Comparisons · Templates · Blog · Docs