Tutorial · Python · ~10 min
How to Upmix Stereo Audio to 5.1 Surround in Python
Beyond matrix upmixing: separate first, then place each source in 3D space. No DSP filter chains, no "fake surround".
Why this works
The principle
Traditional upmixing spreads two channels around the room. AI upmixing splits the mix into sources — vocals, drums, bass, ambience — then renders each source into the surround layout with its own position, depth and gain.
Step 1 — Submit a separation + upmix job
auto placement follows production conventions; manual placement lets you pin each stem yourself.
Python
import mlslabs
client = mlslabs.Client("YOUR_API_KEY")
job = client.source_separation.upmix(
input_url="s3://bucket/concert-stereo.wav",
output_url="s3://bucket/upmix-5-1/",
layout="5.1", # 5.1 | 7.1.2
placement="auto", # auto | manual per-stem
stems=["vocals", "drums", "bass", "guitar", "ambience"],
)
print(job.job_id)Step 2 — Poll and download
Python
job.wait()
print(job.status)
# download: multichannel WAV per layout
print(job.files)Step 3 — Verify the layout
Bash
ffprobe -show_streams upmix-5-1.wav | grep channel_layout
# => 5.1(side) (6 channels)Advanced: manual placement
Pin stems by hand for artistic control — for example, vocals locked to center, rain spread to the rear, thunder routed to LFE and surrounds:
placement=[{"stem": "vocals", "pan": "C"}, {"stem": "rain", "pan": "BL+BR"}, {"stem": "thunder", "pan": "LFE+LS+RS"}]
Note: This is the input side of the stereo-to-Atmos pipeline — the AI Mixing API takes the placed stems and renders object-based immersive masters.