Batch Import

Use Batch Import when you already have audio or video files or want to transcribe direct online network streams in the background.

Best for

• Recorded meetings, lectures, podcasts, or interviews
• Subtitle preparation from local media files
• Workflows where multiple files need to queue and finish one after another
• Direct transcription of online network stream URLs (HTTP/HTTPS/RTMP/RTSP/m3u8) with streaming subtitle generation

Before you queue files

• Configure a Batch Import Model in Settings > Model Settings.
• Confirm your files use a supported audio or video format, or that your network stream URL is accessible.

Add files or streams to the queue

1. Open Batch Import (inside the Create workspace).
2. Drop files into the import area, or click Select File.
3. To transcribe remote audio streams, open the SplitButton dropdown and select Input Stream URL, then enter the stream URL (HTTP/HTTPS/RTMP/RTSP/m3u8).
4. Add one or more items to the queue.
5. Watch the queue sidebar and the active item status. Network stream items show a dedicated Globe badge.
6. Use Add More Files when you want to keep extending the queue.

Adjust import behavior

1. Click Preset Popover Card to adjust Subtitle Mode, Language, or temporary hotwords.
2. Monitor batch progress through the Summary Dashboard (total tasks, total duration, processing, completed, and failed status) and grouped list views.
3. Concurrent Streaming Translation: If translation is enabled, the Streaming Translation Scheduler concurrently translates segments in a sliding window as ASR completes, without waiting for the full stream or file to finish.
4. Real-time Cancellation & Finish & Save: In-flight transcription items can be cancelled in real time directly from the queue or the header Message Center. For active network stream tasks, you can click Finish & Save at any time to immediately finalize already transcribed segments into history.
5. Use the Clear Menu to clear completed, failed, or all items from the queue.

What happens after a file or stream finishes

• Completed items load into the main transcript editor.
• Remote network streams are transcribed in RAM with sliding VAD frame chunking without requiring full disk downloads.
• From there, you can continue with Edit and Playback, AI Polish and Translate, Projects and Inbox, or Export and Settings.

Useful tips

• If no offline batch model is configured, Sona reopens onboarding instead of starting import. Models like Qwen3-ASR (via llama.cpp) or Whisper presets provide high-performance offline batch transcription.
• Settings > Model Settings includes Batch VAD Segmentation (powered by Silero VAD v5), VAD Buffer Size, and Max Concurrent Transcriptions, which affect batch behavior. Turning off batch VAD makes local batch transcription recognize the whole file at once.
• Sona features a built-in pure-Rust audio decoding engine that natively handles WAV, MP3, FLAC, OGG, AAC, and M4A without requiring system FFmpeg installations. If you encounter unusual legacy containers or proprietary codecs, specify a custom FFmpeg fallback path in Settings > Storage Management.
• Batch Import plus export is the main path when you want subtitle files from existing media or remote network streams.