Per-source templates: 808, vocals, drums, master, FX returns. Floating mic is push-to-talk on this tab — toggle modes from the Voice Chat tab.
How to use: click "+ Add plugin" on any bus, search, click to add. ↑/↓ to reorder, × to remove. Hit "Snapshot → Claude" when dialed in — every snapshot is auto-saved as a Session Snapshot under the Voice Chat tab.
Meter Check
Type your live readings — get instant pass/warn/fail vs. the platform target.
Diagnosis
Enter your readings above.
Troubleshooter
Click any symptom for a recipe. Pick multiple — they stack on the snapshot.
Symptoms below feed the same diagnosis Claude sees on the next snapshot.
Drag pills to reorder · × hides a pill · + adds your own symptom
No key saved. Paste and Save.
No key saved. Optional — without it the AI falls back to its built-in (sometimes shaky) knowledge.
OpenAI Realtime
Session$0.000
Balance— set —
Spent$0.000
Left—
Idle
Anthropic Research
Session$0.000
Balance— set —
Spent$0.000
Left—
🛠 Session tools cost & safety
Cost / minute
—
Live during call. Amber over $0.10/min.
Soft budget cap
Voice warning + amber pulse when crossed.
Session breakdown
Audio in$0.000
Audio out$0.000
Text in/out$0.000
Research$0.000
Auto-pause if stuck
Saves you from a runaway session if the model hangs.
Usage dashboards
Real numbers live on the providers' sites.
Live Voice Transcript
Conversation log from your current and most recent voice call.
Transcript will appear here once the call starts.
Session Snapshots
Every snapshot you copy to chat is auto-saved here. Latest first. Tap an entry to expand it.
No snapshots yet — hit Snapshot → Claude to start your session log.
How it works: your API key talks to OpenAI's Realtime API directly from this browser — audio in over WebRTC, audio back live. I've got tools for your chain, library, meters, and flags, so I can actually read and modify the workbench while we talk. Try: "what's on my master bus?", "add Maag EQ4 after Ozone", "switch genre to R&B", or "I'm hearing pumping on the master." Live transcript and session snapshots live just below.
Heads up: voice needs a secure context (HTTPS or localhost). Chrome blocks the mic on file:// URLs — open this artifact via the Cowork sidebar, or serve the folder with python3 -m http.server and open http://localhost:8000. Safari is more forgiving but still needs mic permission.
💬 Live text chat with Claude. Paste session snapshots, ask follow-ups — Claude has your current chain, meters, and full plugin library as context every message.🔊
Mic mode:
Start a chat with Claude. Paste a session snapshot, ask about a producer's technique, or discuss your current chain — Claude has full visibility into your workbench every message.
Quick prompts
Session: $0.000Last reply: —Messages: 0Costs auto-track into the spend counters on the Voice Chat tab (Anthropic for chat, OpenAI for read-aloud).
🧠 Knowledge Base
Paste research from NotebookLM, YouTube transcripts, articles, or interview notes. Active notes are injected into Claude's prompt every chat & voice session.
No notes yet.
Layout
Drag corners to resize · drag a card to move it
Frequency Map
Where each band lives and what to listen for.
Sub-bass
20-60 Hz
808 fundamental
HP below 30Hz if muddy
Bass/Body
60-200 Hz
808 punch, kick weight
Watch masking
Low-Mid (mud)
200-500 Hz
Vocal warmth @ 250Hz
Cut if boxy
Mids
500-2k Hz
Vocal body, snare
Honk ~800Hz
Upper Mids
2-6 kHz
Presence, de-ess zone
Highs
6-10 kHz
Air, cymbal shimmer
Extreme
10-20 kHz
Maag magic band
Loudness Targets
Integrated LUFS by destination.
Competitive Trap
-8 LUFS
Club-loud.
Spotify/YouTube/Tidal
-14 LUFS
Platform normalization.
Apple Music
-16 LUFS
Sound Check.
Broadcast
-23 LUFS
TV/radio (EBU R128).
Pro tip: master to -8, but A/B how it sounds after Spotify normalizes to -14.
True Peak Ceilings
Inter-sample peak limits before encoder distortion.
Streaming
-1.0 dBTP
Safe for MP3/AAC.
SoundCloud
-0.5 dBTP
Stricter encoder.
Safe headroom
-2.0 dBTP
Stems / print masters.
Hard ceiling
0 dBTP
Never exceed.
Stereo Width by Band
How wide each frequency range should image.
Sub-bass
Mono
Below 100Hz mono.
Bass
10-30%
M/S not Haas.
Mids
30-60%
Highs
60-100%
FX Tails
100%
Current genre: Trap— plugins⭐ = top pick for this genre • ❤ = your favorite
Add plugin to bus
Copy manually (⌘A then ⌘C)
Auto-copy failed — your browser blocks clipboard writes on file://. Select all (⌘A / Ctrl-A) then copy (⌘C / Ctrl-C). Fix for good: open the artifact via the Cowork sidebar, or serve the folder with python3 -m http.server in the artifact folder and open http://localhost:8000.
Load a chain preset
💾 Save current chain as a preset
Saves whatever's currently in that bus, with all plugin settings intact.
Load presets into Chain Builder?
Pick which preset loads into each bus. Buses with multiple options for this genre let you choose your favourite — or skip a bus to leave its current chain alone. Closing this dialog without applying keeps your chain unchanged.
✨ Bulk-import plugins via Claude
or paste a list below — or drop a screenshot onto the box / paste an image with Cmd-V
Upload a screenshot or paste a list above, then click Parse with Claude — parsed plugins appear here for review before they're added.