Settings

Status: draft. Concrete reference content + structural placeholders; flesh out the prose + screenshots when ready.

Purpose

The configuration hub. Sections cover:

VL Studio difference — backups and voice

The VL3X has no built-in backup system. The only way to preserve your tuned settings off the device is to take notes, and the only way to migrate to another VL3X is to manually re-create everything.

VL Studio's Backups capture your entire library — presets, songs, setlists, cues, chat sessions, preferences, and AI config — into a single file you can move between machines or restore after a fresh install. Multiple snapshots live side-by-side; selective restore lets you pull back individual sections.

Voice commands (Whisper-based offline speech recognition) and AI Skills (user-editable markdown files that shape how the AI responds) are also entirely new features with no device equivalent.

1. System info + health check — version, library stats, a one-click DB sanity check.
2. Backups — create, browse, restore, download, and import backups; reset preferences and AI config to defaults.
3. AI configuration — provider keys, per-task routing, model overrides, system prompts, custom instructions, token budget, disabled tools, the AI Skills editor.
4. AI Capabilities — read-only list of the tools the AI has access to.
5. Voice commands — Whisper model selection and voice on/off toggle.
6. TALK preset — global slot configuration for the TALK button and talk-break entries.
7. Performance — Advance Cue CC, nav order and hidden-nav-items management.
8. Recordings — folder picker for where recordings are saved.

Walk-through

System info

At the top — app version, preset/system format versions, library counts. Click Health Check to run a quick sanity probe.

Backups

Three actions:

• Create Snapshot — captures the current library state (presets + songs + setlists + chat sessions + preferences) into a backup row.
• Import from File — reads a backup file from disk. Use this to move between machines. You can browse the contents and tick only the sections you want to restore.
• Reset App Preferences / Reset AI Config — restores the bundled defaults.

Existing backups list per-row: Browse (selective restore), Download (save to disk), Restore (full restore — overwrites your current library), Delete.

There's no immutable factory baseline — all backups are equally deletable. To get back to a clean device state, factory-reset the VL3X and run Library → Pull All.

AI configuration

The entry point for everything covered in AI Chat → Reference:

• Provider list — Claude, Gemini, and any number of OpenAI-compatible (OC) providers.
• Per-provider fields — API key, endpoint, model, tool-calling support, timeout.
• Per-provider inference overrides — context window, temperature, max tokens (each can be left as inherit).
• Routing — per-task (generate / describe / suggest / chat) provider selection. Tool-less providers show a "text-only" badge on chat.
• Per-task model override — for Claude / Gemini, pick a specific model per task.
• Custom instructions — global and per-provider text areas appended to the system prompt.
• System prompts — per-task preview, read-only by default; unlock for editing behind a confirmation modal. Reset to default is always available.
• Token Budget % — how much of the provider's memory window is reserved for chat history.
• Disabled Tools — per-tool checkboxes; off = hidden from the AI.
• AI Skills editor — per-skill toggle, edit modal (name / triggers / priority / body), import / export.

AI Capabilities

Read-only list of the tools the AI has, grouped by purpose (Sound Design / Device Control / Songs & Setlists / Undo). Useful for verifying what the AI sees vs what you've disabled.

Voice commands

• Model selection — three tiers (Fast / Balanced / Accurate). Download from Hugging Face, mark active, delete from disk. A small bundled model is installed automatically on first launch so voice works without internet.
• Voice toggle — turn off to unload the voice model from memory entirely. Frees roughly 150–500 MB depending on tier. Re-enable to reload it when you next use voice.

TALK preset

Numeric input for the global TALK preset slot. Default suggested value: 490 (TC-Helicon factory blank). Live preview shows the preset name at the configured slot. If you leave it empty, per-cue or per-setlist-entry overrides still work.

The TALK slot is shared by:

• The Device page TALK button.
• The Performer Mode header TALK button.
• Setlist talk-break entries without a per-entry override.
• Song cue TALK actions without a per-cue override.

Performance

• Advance Cue CC — assign a CC number to "advance one cue" for Performer Mode. Useful if you have a footswitch you'd rather not have changing presets.
• Nav order / hidden nav items — drag-reorder the main nav tabs, with per-tab visibility toggle. Hide pages you don't use.

Recordings

Folder picker for where audio recordings get saved. Default is your platform's Music folder (~/Music/VL Studio/ on Linux). The folder is created if it doesn't exist when you start recording.

Reference

Backup format

Backups are gzipped JSON (.vl3x-backup.json.gz) containing your presets, songs, setlists, cues, chat sessions, preferences, and AI config. They round-trip across machines. Restoring is selective per section.

Note on restored presets: the SysEx data is recreated from each preset's stored parameters rather than copied verbatim. Backups are data, not binary device-state archives — if you ever need a true bit-for-bit copy of what's on the device, pull from the device itself.

For the full interchange story — songs, setlists, presets, backups, and AI Skills as standalone files — see Sharing & Files.

Troubleshooting

Symptom	Fix
Reset App Preferences / Reset AI Config seems to do nothing	The confirmation modal must be accepted. Click through it — the reset writes the bundled defaults.
Yellow "Custom prompt active" banner won't clear	A per-task AI system prompt has been overridden. Settings → AI → System Prompts → click Reset to Default on the highlighted task.
TALK preset preview says "no preset at this slot" but a preset exists on the device	The preview reads your local library. If your library is empty (fresh install, no Pull All yet), the message fires even for slots that exist on the device. Run Library → Pull All.
Voice toggle off → on doesn't reload the model immediately	The model loads on demand the next time you use voice. Send a voice command after toggling on.
Advance Cue CC fires twice per footswitch press	The card debounces presses, but a device latch mode can still emit two presses. Check the footswitch latch mode on the Mixer page.

See also

• AI Chat — uses the AI configuration set up here.
• Audio — uses the recordings folder set up here.
• Performer Mode — uses the TALK preset slot and Advance Cue CC set up here.
• Songs — TALK cue actions fall back to the global TALK preset slot from here.
• Sharing & Files — full backup format, plus all the per-item exports that live alongside it.