SFUMIA / Documentation
Guide
Mix dialogue, music and effects
Put dialogue and music on separate tracks.
In this guide
Example
Set clip levels first, then track and master levels. A group collects tracks processed together. Check without Solo and measure the final file; speaker volume does not change exported levels.
Adjust picture and sound
The mixer separates clip, track, group and master output. Check routing and levels; turn off solo before checking the full film.
Audio effects and new takes
Depending on the clip, use gain, balance, fades, EQ, compressor or limiter. The output limiter is not a certified true peak meter: check the final file against your delivery requirements.
Create Sound prepares a new take. Listen to it, then insert it explicitly. Installing a model, generating and inserting are separate actions; a local error does not start paid Cloud processing.
Mixer and limiter
The mixer separates the tracks, the « Voix », « Musique » and « Bruitages » groups and the « Sortie générale ». Each strip offers Mute, Solo (except the master output), « Gain (dB) » from −60 to 12 and « Balance ». « Protection de la sortie générale » enables a limiter with « Plafond (dBFS) » from −12 to 0 and « Relâchement (ms) » from 10 to 1,000; it acts on samples, not on peaks between samples.
« Volume dans le temps » places up to 64 points per clip, from −60 to 12 dB; the points follow cuts and speed. Solos and mutes are part of the exported mix; at most 128 audio voices play at the same time. « Export audio », in the mixer, produces a 48 kHz 32-bit float stereo WAV of the full mix, of one track or of one group.
Play video and listen to audio
Use play, pause, frame stepping, in/out points and looping. With the Viewer active, J, K and L control transport. The cache bar shows loaded frames; missing frames remain identified.
For compatible audio, adjust listening controls and use the waveform to navigate. Audio mixing and export take place in Montage.
On-demand audio measurements
Channel analysis and LUFS/true-peak analysis inspect authorised media without changing the file or listening volume. They require compatible audio components in the installation.
A mono waveform does not show every channel difference. Silence, very short audio or interrupted analysis may not provide a usable measurement.
What the analyses measure
« Mesurer les canaux » calculates, for each channel of the first audio stream, the sample peak, the RMS level in dBFS, the DC offset and the number of full-scale samples; up to 32 channels, 384 kHz and 24 hours. Sample peaks are not true peaks and the RMS is not weighted.
« Mesurer LUFS / crête réelle » gives the integrated loudness, the loudness range (LRA) and the true peak, to two decimal places, up to eight channels (7.1) and 192 kHz, without dual-mono compensation. Below 400 ms, integrated loudness is not measured; LRA needs at least 3 seconds.
Save your work
- Check source warnings and the node you are editing.
- Use Save or Ctrl + S and wait for confirmation.
- Create a version or publish through the node controls only when you choose to.
Available features depend on your version and installation.