Recording
Press REC, and the video already has the sound
Three ways out of the stage: an MP4 with H.264 video and AAC audio in one file, an AVI with a WAV beside it, or a 32-bit PNG sequence with real transparency for compositing. Resolution and frame rate are yours, and the recording is drawn at its own size — not upscaled from the preview.
Formats
Which one to pick, and what your PC needs
| Format | What you get | What the PC needs | Best for |
|---|---|---|---|
| MP4 (H.264 + AAC) | One file with the picture and the sound already inside it; each audio sample carries its own timestamp, so nothing drifts. | Windows' own Media Foundation encoders — H.264 and AAC. Nothing to install. | YouTube, TikTok, sending a take to somebody. |
| AVI video | The picture in one file plus a 16-bit stereo WAV beside it (take.wav), and the exact ffmpeg line printed for you after you stop. | An MJPEG codec (for example the K-Lite pack) if you want smaller files — optional; without one the AVI is uncompressed RGB and stops at the 2 GB AVI limit. | Editing suites that want an intermediate file. |
| PNG sequence (32-bit alpha) | A folder of frame-000001.png files with a real alpha channel, plus sequence.json recording the size, the frame rate, the frame count and the ffmpeg command that rebuilds an alpha video. | Nothing extra; this path is drawn by the software renderer so transparency is exact. | Compositing over your own footage in Premiere, Resolve or OBS. |
Format, resolution, frame rate, the audio track and transparency live in the OUTPUT card of the Recording page and apply from the next take.
MP4 in one file
Written by Windows itself — no ffmpeg, no extra library
- A Media Foundation sink writer takes the stage frames, converts them to NV12 (BT.601) and encodes H.264 with a bitrate computed from size × fps — a 2 Mbps floor and a 24 Mbps ceiling.
- The engine's PCM is encoded to AAC into the same file, and every sample keeps its own timestamp, so picture and sound cannot drift apart.
- A machine with no H.264 encoder gets a clear sentence with the error code instead of a broken file; one with H.264 but no AAC encoder still records the picture, and the dialog says why the sound is missing.
- No SoundFont loaded means picture only, and the status line says so before you press REC.
The audio track
The same samples you hear, captured while you play
- Record audio is on by default. The tap takes the very PCM blocks the render engine plays, and it is attached only while a recording runs — performing normally costs nothing.
- AVI and PNG sequences get a 16-bit stereo 44.1 kHz WAV beside the video: take.wav, or audio.wav inside the frame folder.
- A machine with no audio output device is not a problem: the engine renders the audio itself, so the take still has its full length.
- After you stop, the dialog prints the length and the ffmpeg line that muxes the two files into one MP4.
- The info line adapts to the machine: with a SoundFont it says where the sound goes, without one it says plainly that the take will be silent.
The line the dialog prints
ffmpeg -i video.avi -i take.wav \
-c:v copy -c:a aac take.mp4Copy it, run it, and the AVI plus its WAV become one MP4 — the video stream is copied, not re-encoded.
Recording from the GPU stage
An extra off-screen frame, at exactly the recording's size
While the GPU stage is up, REC makes the render loop draw one more frame off-screen at exactly the resolution and frame rate of the recording — not an upscaled copy of the preview window. The WPF layers (the staff, the camera picture, the counters) are composited on top of it.
- The render loop alone — no window, no preview — still hands over exact-size frames, which is what the built-in verification suite checks.
- Transparent PNG keeps using the software renderer, because that is the path that can produce a real alpha channel.
- The Green Screen preset paints a pure green backdrop, which is exactly the colour the camera overlay keys out — so OBS can key the stage too.
A take
From preset to published video
- Set the outputRecording → OUTPUT: format, resolution (window / 720p / 1080p), frame rate 15–60, Record audio, and Transparent background for PNG.
- Choose the lookPick a preset in Style, an interface in Theme, and load the score you want to play.
- Press RECREC sits in the corner of the stage: press it, choose where the take goes, play, then press it again to stop.
- Take the fileThe dialog reports the length and, for AVI or PNG, prints the ffmpeg line that finishes the job.
After the take
Two ffmpeg lines you will actually use
# AVI + WAV -> one MP4 (video copied, audio encoded) ffmpeg -i video.avi -i take.wav -c:v copy -c:a aac take.mp4 # PNG sequence with alpha -> WebM with alpha ffmpeg -framerate 60 -i frame-%06d.png -c:v libvpx-vp9 -pix_fmt yuva420p take.webm
sequence.json, written next to the frames, carries the size, the frame rate, the frame count and a suggested command for exactly this.
What recording cannot do yet
There is no direct WebM/VP9 writer — the 32-bit PNG sequence is the path to alpha video, and sequence.json suggests the ffmpeg line that builds it. MP4 needs Windows' own encoders, so a machine without H.264 cannot write that format at all, and a machine without AAC records the picture only.
Your next video is one key away
REC is on the stage from the first launch — MP4 with the sound already inside it.