SPEC Shorthand Codec (Beta)

Lossless ~-token encoding you can decode deterministically. Semantic mode may apply lossy vowel reduction. Token savings vary by tokenizer.

Try Prompt Compactor Join Waitlist

Download dictionary JSON (for your own integrations).

Live SPEC Demo

Original Text

0 chars

SPEC Output

0 chars
0%
Reduction
0
Chars Saved
—
Round-trip
0ms
Processing Time
—
Dict Entries

Try These Examples

Prompt Compression

"Please summarize this information and generate a short response for the assistant."

~33% reduction (lossless)

Phrase Compression

"in other words, for example, explain the system configuration."

~45% reduction (lossless)

Case Flags

"Tomorrow INFORMATION: please review and update."

~26% reduction (lossless)

Semantic Mode

"characterization and interoperability are important for documentation."

~39% reduction (semantic)

How SPEC Works

1

Lossless Codes

Dictionary words become short ~-prefixed tokens (safe to decode)

the → ~t, because → ~bc
2

Phrase Codes

Common prompt phrases compress to short codes (lossless; tokenizer-dependent)

for example → ~p/eg, in other words → ~p/iow
3

Case + Escapes

Preserve case and prevent accidental expansions

Tomorrow → ~^tmrw, ~ → ~~
4

Semantic Mode

Optional vowel compression for extra savings on long lowercase words (not perfectly reversible; skips mixed-case identifiers)

characterization → charctrztn