Recording
Full-session multi-stem recording, highlight markers, and instant-replay clipping in DualStream.
Recording in DualStream covers three independent workflows: full-session multi-stem recording, highlight markers for later editing, and instant-replay clipping from a rolling buffer. All three run alongside streaming and share the same GPU-composited feed, so recording never re-renders or re-encodes what's already going out.
Full-Session Recording
Full-session recording captures your entire broadcast to disk as separate, remixable tracks. Open the Record tab in the main sidebar to access the recording panel.
Recording Panel
The Record tab has two states driven by whether a past session is selected:
- Idle state — capture controls at the top, followed by a session list grouped by Today / Yesterday / older. Click any session to browse its tracks.
- Session-selected state — compact capture controls, a Back action, Open folder, and a per-track list with mute toggles for playback preview.
While a recording is active the panel also shows:
- A live file list with per-track names and sizes that update as the session grows.
- The Highlight Markers list (see below), populated as you mark moments.
Starting and Stopping
Press F8 (default hotkey) or click the Start button in the Record tab. Press again to stop. The toolbar at the top of the app shows a red-dot indicator and elapsed timer while recording.
Stems
DualStream's recorder is multi-stem by design. Every session produces:
| Stem Type | File(s) | What It Contains |
|---|---|---|
composite_av | stream_h.mkv, stream_v.mkv | Full composite A/V for the horizontal (16:9) and vertical (9:16) canvases — exactly what viewers see. |
mixed_audio | mixed_audio.mkv | Master audio mix as a standalone audio-only file, ideal for podcasts or audio-only uploads. |
audio_stem | <source>.mkv per audio source | Every audio source recorded to its own file — mic, game, music, browser sources, etc. |
Per-source audio stems let you remix in post, swap out copyrighted music before uploading to YouTube, or mute individual sources without re-recording. All files are written to a single session directory and can be loaded back into any DAW or NLE.
Highlight Markers
While recording, press F11 (default hotkey) or click the Highlight button to drop a timestamped marker at the current position. Markers appear in the panel's Highlights list with their MM:SS offset and an optional label, and are saved alongside the session so you can jump straight to them when editing.
Markers don't create separate clips — they annotate the full-session recording.
Instant-Replay Clipping
The clip feature maintains a rolling buffer in memory so you can save the last 30, 45, 60, or 90 seconds of your stream on demand, without recording the full session.
- Enable the clip buffer before or during your stream.
- When something clip-worthy happens, press the clip hotkey (default Ctrl+Shift+C) or click the Clip button in the toolbar.
- DualStream writes both the horizontal and vertical versions of the clip to disk.
The clip buffer indicator shows when buffering is active and how much footage is available. Clipping is independent of full-session recording — you can use either, both, or neither.
Clip Settings
- Buffer Duration — 30, 45, 60, or 90 seconds of footage retained in GPU memory.
- Default Clip Duration — 10, 15, 30 seconds, or full buffer when saving.
- Auto-start on stream — Begin buffering automatically when you go live.
- Max clips to keep — Optional rolling cap on saved clips.
- Keyboard shortcut — Default
Ctrl+Shift+C, remappable in settings.
Settings
Settings → Recording controls the full-session recorder:
| Setting | Default | Notes |
|---|---|---|
| Output directory | Videos\DualStream\Recordings | Each session is written to a timestamped subfolder. |
| Mixed-audio bitrate | 256 kbps (AAC) | Used for composite_av audio tracks and mixed_audio. |
| Audio-stem bitrate | 192 kbps (AAC) | Used for per-source audio stems. |
| Auto-start on stream | Off | When on, recording starts automatically as soon as you go live. |
| Auto-stop after | Unlimited | Optional session cap in hours. |
| Split files every N hours | Off | When set, uses GStreamer splitmuxsink to roll files — useful for all-day sessions. |
| Low-disk-space warning | 10 GB | Shows a warning banner when free space falls below this threshold. |
| Toggle hotkey | F8 | Start/stop recording. |
| Highlight hotkey | F11 | Mark a highlight during recording. |
FAQ
What's the difference between full-session recording and clipping?
Full-session recording writes your entire broadcast to disk as separate stems you can edit later. Instant-replay clipping saves short, fixed-length highlights from a rolling buffer without keeping the whole session. Use full-session for editing and archiving; use clipping for social media and quick shares.
Can I record while streaming?
Yes. Recording runs on a branch of the same GPU-composited pipeline that drives the stream, so there's no second encode pass and minimal additional CPU or GPU cost.
Why are recordings saved as .mkv instead of .mp4?
Matroska (MKV) tolerates crashes and power loss — if DualStream is killed mid-recording, the file is still playable. MP4 requires a clean finalize step that can lose the whole file on crash. MKV also handles multi-track audio more cleanly, which matters for stem recording. You can remux to MP4 in any editor or with ffmpeg if you need it.
How do I strip copyrighted music before uploading?
Start a full-session recording with your music on its own audio source. After the session, mute the music stem and re-mix the remaining stems in your editor — the composite stream_h.mkv / stream_v.mkv gives you the visual reference while the individual audio files give you the flexibility.
Do highlight markers create separate clips?
No — markers are metadata on a full-session recording, not cuts. Use them as jump-points when editing the full session in your NLE, or combine them with instant-replay clipping if you want standalone files.
Does recording affect stream performance?
The recording pipeline shares the GPU composite with the stream encoder, so the overhead is limited to the extra file encoders and disk writes. Expect a small CPU increase proportional to the number of stems enabled.
Last updated on