Getting Started

1. Download & Install

Visit the homepage to download the installer for your platform:

• macOS: Download .dmg, drag to Applications
• Windows: Download .exe installer, follow the wizard

Requirements: macOS 11+ / Windows 10+, 8GB+ RAM recommended.

2. First Launch & Model Download

On first launch, if no transcription model is detected, you’ll be guided to download one:

1. Choose an ASR engine:
   • Whisper: strong multilingual support; we recommend base for beginners
   • Qwen3-ASR: better for Chinese, dialects, and Chinese-English mixing; we recommend 0.6B for beginners
2. Choose a default model
3. Click download — the model will be saved locally
4. Enter the workbench after download completes

You can also import custom models via Deep Link:

oncescribe://download/?url=<model_url>

3. Complete Your First Transcription

Method 1: Local File

1. Select the “File” tab on the workbench
2. Drag & drop an audio/video file, or click to select
3. Click “Transcribe”
4. Wait for completion — results appear on the right with timestamps

Method 2: Recording

1. Select the “Record” tab
2. Choose input device (microphone / system audio)
3. Click “Start Recording”, stop when done
4. Recording auto-saves and transcribes

Method 3: Link

1. Select the “Link” tab
2. Paste a YouTube or similar video URL
3. Install yt-dlp on first use (guided)
4. Auto-transcribes after download

4. Export Results

After transcription, click the export button in the toolbar:

• TXT: Plain text, good for reading
• SRT/VTT: Subtitle files for video production
• JSON: Full data with timestamps
• DOCX/Markdown: Rich format documents

5. Explore More

• Transcription Settings — Adjust model, language, temperature
• AI Summary — Generate content summaries with LLM
• Batch Processing — Process multiple files at once
• Speaker Diarization — Enable diarization to identify speakers

Need help? Check FAQs or email contact@once.work