● Knowledge Base · Panels

AI Agent Panel

The AI workbench inside basehead — three tabs (Ask, Scene Builder and Skills) from asking operational questions all the way to image and audio upload and analysis.

Ask All editions

Ask basehead questions about the program, your library, or how to do a thing. Open it from View → AI Agent or F12.

This is a more advanced T-2000 chatbot that will show you pill buttons that will take you to items or run functions that you were looking for directly in the program.

You get 5 free questions per day on basehead‘s dime — add your Anthropic API key in Options → API Keys to get unlimited access.

Scene Builder Ultra

Feed it a reference image and Scene Builder turns it into search terms (or UCS categories) for matching sounds in your library.

Loading a reference image

Click the box to paste, drag & drop a PNG/JPEG/WebP/GIF, hit Choose to open a file picker, or Ctrl/Cmd + V.

Output

• Search Terms — free-form sound search terms (with the AMB/SFX/Foley filters, tone, time-of-day).
• Categories — top-level UCS Category only (e.g. AMBIENCE).
• Categories + Subcats — full UCS pair (e.g. AMBIENCE - WIND).

Both category modes stay inside the real UCS v8.2 taxonomy — no invented names.

Filters

AMB / SFX / FOLY pills are radio — pick one. When AMB is picked and Output is Search Terms, an extra EXT / INT row appears to steer toward outdoor or indoor ambiences.

Then Build from the suggestions

Once the results are how you like them (delete the duds, keep the keepers), the Create button at the bottom of the panel is a split button with a ▾ dropdown — three ways to put those terms to work (only available when Output is set to Categories or Categories + Subcats):

• Tab Cluster — opens a new search tab for each category, then groups them under one cluster so you can recall the whole working set with one click.
• Collection — drops every matching record into a new Collection in the NodeTree.
• Taglists — creates a Taglist for each category so you can audition / spot from them.

Skills Ultra

Bulk operations the AI runs across the records you’ve selected. The Skill dropdown picks which one:

1. Tag Creator — listens to audio (or reads filenames / descriptions) and writes back tags + descriptions.
2. Category Matcher — classifies records into a UCS Category / Subcategory.
3. SQL Commander — describe an edit in plain English, get a scoped SQLite statement to review and Apply.
4. Language Translator — translate a field into another language.
5. The Cleaner — tidy spacing, punctuation, spelling, grammar.

---

Tag Creator

Drop a selection of audio files on Tag Creator and it returns SHORT + LONG descriptions and SFX TAGS for each, written back to the metadata fields you choose.

Input

Three sources to pick from — Audio, Filename, Description.

(a Gemini API Key is needed for audio analysis)

Output routing — TAGS / SHORT / LONG

Each result has its own destination dropdown: category, comments, description, filename, keywords, notes, subcategory, track_title.

---

Category Matcher

Classifies selected records into a UCS Category / Subcategory from the official UCS v8.2 taxonomy. Suggestions are constrained to the real UCS list — no invented names.

Source & output

The Source (combine fields) picker decides what the AI reads. description is the default, but you can chain multiple fields together for more context (e.g. description + comments). The skill then writes the chosen UCS pair back to the record’s category and subcategory columns.

Options

• Add CatID to [filename] — prefixes the UCS CatID to the filename (destructive — same warning as the Tag Creator filename routing).
• Burn Into Metadata — also writes the UCS fields into the audio file’s iXML metadata, not just the basehead database.

Run on a selection — the result list shows the classified pair per record (e.g. WINGS / CREATURE) so you can spot-check before relying on it for a bulk move or rename.

---

SQL Commander

Also known a Batch Renamer on Crack! Describe the edit in plain English and SQL Commander returns a scoped SQLite statement that touches only the records you’ve selected. Review the SQL, then Apply. This is beyond a powerhouse feature and it let’s you edit metadata in an entirely new way with ease.

Safeguarded vs Unrestricted

• Safeguarded (default) — only writes to the curated list of safe-to-edit columns.
• Unrestricted — lifts the field-write restriction so you can edit fields the curated list omits (numeric, date, user1-8). All other safety rails stay.

Schema awareness

Custom fields user1 – user8 are sent to the AI with their configured display names. “Clear the first four user fields” maps to user1…user4. A custom field referenced by its display name resolves correctly. If the AI can’t map part of a request to a real column, it declines instead of guessing.

---

Language Translator

Source field + target language + destination field. Translated text writes to the destination.

Only translate known words

Default OFF. When ON, drops any term that can’t be translated — proper nouns, brand / mic names, codes, abbreviations. The destination ends up as a clean list of translated words, intended for building foreign-language search keywords alongside the originals.

---

The Cleaner

The Wolf is here to tidy your metadata — auto-fixes spacing, punctuation, spelling and grammar — never rewrites or adds words. AI-only — for non-AI tidy use the Batch Renamer.

Options

• Field picker — which field to clean.
• Capitalization — Leave as-is · First letter · Title Case · lower case · UPPER CASE.
• Add ending period — for description / comments / notes.
• De-Fluff — condenses prose into terse sound-library style. The only mode that may reword text.

Preview → Inject

• Preview — shows before → after. Nothing is written.
• Clean — writes the tidied fields to the database.
• Clean and BurnIn — writes the database AND burns the changes into the audio files.

After a Preview, an Inject split button appears — Inject commits to the database. Inject and Burn commits to the database and files.

FAQ