Search Modes
The Sample Browser’s search bar has three modes that change how your query is interpreted. Switch between them with the toggle above the search bar, or pressTab while the input is focused. Each panel remembers its own mode in your dock layout. The ? button at the end of the search bar opens a quick reference with the same info.
| Mode | How it works | Trigger | Cost |
|---|---|---|---|
| Find (default) | Filters by filename and tokens | Live as you type | Free |
| Prompt | AI semantic search — describe any sound in plain English | Enter or Search button | 5 credits |
| Instant | On-device matching using a local model — fast but less accurate than Prompt | Enter or Search button | Free |
- Find is best when you know the sample name or want to scope by tokens. Fully offline.
- Prompt is best when you’re searching by vibe or mood, especially across poorly-named samples. Requires internet.
- Instant is best for quick exploratory searches when you don’t want to spend credits. Fully offline.
When credits are unavailable, Prompt mode falls back to a local model
(equivalent to Instant). Free plan users receive 100 daily credits — enough
for about 20 Prompt searches per day.
Writing effective prompts
Prompt and Instant both match on how the sound feels, not on exact specs. Describe the character, role, and vibe you’re after and let the filters handle the numbers.- Mood and character:
dark and atmospheric,warm and nostalgic,aggressive and gritty - Instrument or role:
trap hi-hats,analog bass line,ambient pad for a breakdown - Production style:
lo-fi dusty drums,vintage tape saturation,clean modern synth - Adjectives that describe timbre: warm, punchy, smooth, gritty, airy, metallic
Search suggestions
Before you type, you’ll see suggestion chips personalized to your search history. Click any suggestion to run that search immediately. Suggestions adapt over time as you search more.Combining tokens with prompts
All three modes support the same token grammar —#120, @Cmin, !loop, $techno, %dark. In Prompt and Instant, tokens become hard filters on top of the AI match, so only the plain-text portion drives the semantic search. See Search Syntax for the full token reference.