A workflow for using Phantom as a measurement companion during mixing. Instead of guessing whether your EQ moves worked or your compression is appropriate, measure the results and iterate with data.
When to use this
- During active mixing sessions when you want verification
- After making EQ/compression moves — “did that actually help?”
- When your ears are fatigued and you need objective data
- Checking mix balance between sections (verse vs chorus)
Workflow Overview
Set measurement checkpoints
Capture the state before processing so you can measure improvement
EQ with masking data
Use Phantom's masking analysis to guide surgical EQ decisions
Verify dynamics after compression
Confirm you haven't over-compressed by checking crest factor and range
Check stereo image and phase
Ensure your panning and spatial processing translate to mono
Compare mix versions
A/B your current mix against previous versions with actual measurements
Step 1: Set measurement checkpoints
Why: Before making any processing changes, capture baseline measurements. This lets you objectively compare “before” and “after” instead of relying on memory (which isn’t reliable after 30 minutes of listening).
Prompt:
Run a full diagnostic on my current mix bus at mix-v1-unprocessed.wav — I’m about to start processing and want a baseline
Phantom output:
Full Diagnostic: mix-v1-unprocessed.wav Duration: 3:42 | 44.1kHz/24bit | Stereo
Spectral Centroid: 2,456 Hz Rolloff: 10.8 kHz
Loudness Integrated: -18.2 LUFS True peak: -4.1 dBTP Range: 9.8 LU
Dynamics Crest factor: 14.2 dB Complexity: 0.81
Stereo Width: 0.58 Correlation: 0.72
Problems MOD Mud buildup 200-400 Hz LOW Sub energy below 30 Hz (rumble)
What to note: Save these numbers (or let your AI assistant remember them). After processing, you’ll compare back to this baseline to verify improvement.
Step 2: EQ with masking data
Why: Instead of sweeping EQ blindly, use masking analysis to know exactly where to cut. You’re making informed surgical decisions, not exploratory ones.
Prompt:
I’m about to EQ the guitars to make room for vocals. Show me exactly where they’re conflicting so I know where to cut
Phantom output:
Masking Analysis: vocals.wav masked by guitar-bus.wav
Overall masking severity: 0.58 (moderate)
Per-octave Overlap: 250 Hz: 0.34 mild 500 Hz: 0.61 significant 1 kHz: 0.72 significant 2 kHz: 0.54 moderate 4 kHz: 0.28 mild
Primary conflict zone: 500 Hz - 2 kHz Recommendation: Cut guitar 2-3 dB at 800 Hz - 1.5 kHz (Q=1.0)
What to do: Apply the suggested EQ cut on the guitar bus, then re-run the analysis.
If you’d rather apply the cut directly as a processed file, use apply_processing:
Prompt:
Apply a 2.5 dB cut at 1 kHz (Q=1.0) to guitar-bus.wav and save it as guitar-bus-eq.wav
apply_processing lets you define EQ curves, high/low-pass filters, and compression settings as a chain. It writes the result to a new file so your originals stay intact. Requires pip install phantom-audio[processing].
Then verify the masking improved:
Prompt (after EQ):
I’ve cut the guitars at 1 kHz. Re-check the masking between vocals and guitars — did it help?
You should see the masking severity drop in the conflict zone. If it’s still above 0.5, you may need a wider cut or a different approach (volume automation, panning separation).
Pro tip
A 2-3 dB cut that’s informed by measurement data is worth more than a 6 dB cut based on guessing. Small, precise moves preserve the tone of both elements while creating separation.
Step 3: Verify dynamics after compression
Why: Compression is the most common cause of “my mix lost its punch.” Measuring crest factor before and after tells you objectively how much transient impact you’ve sacrificed.
Prompt:
Compare the dynamics of drums-raw.wav versus drums-compressed.wav — did I kill the transients?
Phantom output:
Dynamics Comparison:
drums-raw.wav: Crest factor: 15.8 dB Dynamic range: 18.2 dB Complexity: 0.74
drums-compressed.wav: Crest factor: 9.2 dB Dynamic range: 11.4 dB Complexity: 0.52
Change: Crest: -6.6 dB (significant transient reduction) Range: -6.8 dB Assessment: Heavy compression applied. Transient character significantly altered. Consider parallel compression or lighter ratio to preserve more punch.
Decision point: A crest factor drop of 3-4 dB is typical for moderate compression. A drop of 6+ dB means you’ve significantly changed the character. Decide if that’s what you wanted.
Step 4: Check stereo image and phase
Why: After panning, stereo effects, and bus processing, verify that your mix translates to mono and doesn’t have unexpected imbalances.
Prompt:
Check the stereo image of my mix bus — is it balanced and mono-compatible?
Phantom output:
Stereo Analysis: mix-v2.wav Width: 0.74 Balance: -0.06 (very slightly left-heavy) Mid/Side: 58% / 42% Correlation: 0.56
Phase Analysis: Overall coherence: 0.71 Per-band: Sub: 0.96 (solid) Low: 0.88 (good) Mid: 0.68 (moderate) High: 0.48 (wide but risky)
Note: High-frequency coherence below 0.5 may cause brightness loss in mono playback.
What to look for:
- Balance should be within +/- 0.05 of center. A -0.06 left lean is borderline — investigate
- Correlation below 0.5 in any band means mono cancellation risk
- High-frequency coherence at 0.48 suggests stereo widening effects may be too aggressive
Pro tip
If high-frequency coherence is below 0.5 and you’re using stereo widening, try reducing the width effect above 8 kHz. Most stereo enhancement works best in the mid-range (1-8 kHz) where it creates space without risking mono collapse.
Step 5: Compare mix versions
Why: After a mixing session, compare your processed mix against the unprocessed baseline from Step 1. Objective data shows what improved and what might have gotten worse.
Prompt:
Compare mix-v2-processed.wav against my baseline mix-v1-unprocessed.wav — what improved and what got worse?
Your AI assistant will compare the measurements and identify:
- Did the mud problem from the baseline get resolved?
- Did loudness/dynamics change appropriately?
- Is the stereo image better defined?
- Were any new problems introduced?
Quick Reference
Run a full diagnostic on [mix bus] -- I want a baselineShow me where [element A] and [element B] are conflictingCompare dynamics of [raw] vs [processed] -- did I overdo it?Check the stereo image of my mix bus -- is it balanced and mono-compatible?Compare [new version] against [baseline] -- what improved?
Next Steps
- Mastering prep workflow — When your mix is done, prepare for mastering
- analyze_masking — Deep dive on masking analysis
- analyze_stereo — Stereo image measurement details