The Imagery & 3D Suite

Ten codecs, engineered from first principles.

Adopt a Bitruvius flagship codec, or swap a library you already ship for a faster one and change nothing else. Every codec in the suite is memory-safe, deterministic across architectures, and wire-format compatible end to end.

Bitruvius flagship codecs

RIPT

Scientific raster.
Imagery & elevation.

Ours · patent-pending

BVC

Point clouds
& Gaussian splats.

Ours · patent-pending

Drop-in replacements

TurboLERC replaces Esri LERC TurboWebP replaces libwebp 1.6.0 TurboJXL replaces libjxl 0.11.2 TurboLZW replaces libtiff 4.7.1 LZW TurboLZ4 replaces liblz4 (official C LZ4) TurboZstd replaces libzstd (RFC 8478) TurboLEPCC replaces Esri liblepcc TurboSPZ replaces Niantic libspz v4

Keep your format, swap the backend, rewrite nothing. Every one is memory-safe, which the C originals are not.

10

codecs, one vendor

8

drop-in replacements

2-6

weeks to pilot, not quarters

0

lines of code rewritten

The Suite

Pick your level.

Track B headline speedups

Encode and decode wins vs each upstream library on the Bitruvius 804-sample corpus. Hover a row for context; click for full product detail.

Encode Decode ▲Parallel peak — — Upstream parity
0×1×2×3×4×5×6×7×8×Speedup vs upstream library1× parityTurboLERCvs Esri LERC2.6× enc1.9× decwire-identicalTurboWebPvs libwebp 1.6.04.5× enc1.2× dec4.4% largerTurboJXLvs libjxl 0.11.2▲ per-core encode vs cjxl-e7 71.0×2.0× enc1.0× dec3.3% smallerTurboLZWvs libtiff 4.7.1 LZW▲ 24-core parallel 10.0×1.2× enc1.2× decwire-identicalTurboLZ4vs liblz4 (official C LZ4)▲ RGB ortho decode 3.5×1.5× enc2.0× dec0.5% largerTurboZstdvs libzstd (RFC 8478)7.5× enc1.4× dec2.3% smallerTurboLEPCCvs Esri liblepcc▲ Intensity dec (NEON) 4.1×1.8× enc1.9× decwire-identicalTurboSPZvs Niantic libspz v4▲ multicore encode 7.3×4.2× enc1.4× dec3.3% smaller
TurboLERC replacesEsri LERC Mac NEON 2.64x enc / 1.91x dec library swap (turbolerc-libwebp-shim / TIFFRegisterCODEC path varies) TurboWebP replaceslibwebp 1.6.0 4.5x encode, 27/32 decode cells beat libwebp; within 4% on size turbowebp-libwebp-shim cdylib (LD_PRELOAD / DYLD_INSERT_LIBRARIES / DLL drop) TurboJXL replaceslibjxl 0.11.2 smaller than cjxl -e1, memory-safe, ~42-71x faster per-core encode vs cjxl -e7 cjxl/djxl-compatible CLI + native API (bit-exact wire) TurboLZW replaceslibtiff 4.7.1 LZW 1.15-1.20x Intel, +33-35% NEON, 10x parallel C-ABI cdylib via TIFFRegisterCODEC; LD_PRELOAD / DLL drop TurboLZ4 replacesliblz4 (official C LZ4) 2.05x decode, 1.49x fast-encode, size parity lz4-compatible native API; block + .lz4 frame wire TurboZstd replaceslibzstd (RFC 8478) 1.7-3.1% smaller L1-L9, 5-10x archival zstd-compatible native API; RFC 8478 wire TurboLEPCC replacesEsri liblepcc Beats the Esri C++ reference on every kernel; 9.74x peak Native library / FFI (ArcGIS I3S consumer) TurboSPZ replacesNiantic libspz v4 4-7x encode; archive profile 3.1-3.5% smaller AND ~1.6x faster; bit-exact libspz v4 quantization turbospz-ffi cdylib + cbindgen header; native library (turbospz-safe)

Tiered adoption economics

Value compounds with format coverage.

The largest structural win comes from adopting a Bitruvius flagship codec (RIPT for raster, BVC for volumetric) on greenfield and migrate-able workflows. The interim tiers grow with the number and types of imagery formats you swap onto the Bitruvius drop-ins.

TierAdoption shapeCumulative multiplierCompounding drivers

Tier 0

Single drop-in pilot

1 Turbo* product replacing 1 upstream1.0xBaseline - in-product storage / bandwidth / CPU savings only

Tier 1

Multi-format drop-in adoption

2-3 Turbo* drop-ins live1.5-2.0x+ single-vendor procurement, shared support contract, cross-product engineering integration

Tier 2

Full Turbo* suite adoption

4-6 Turbo* drop-ins live2.5-3.0x+ cross-format pipeline efficiency, unified observability, consolidated security review (one zero-unsafe posture, one CVE attack surface)

Tier 3

RIPT / BVC native-format adoption

RIPT (raster) or BVC (volumetric) replaces the native format for greenfield or migrate-able workflows4-6x+ structural compression wins not available in upstreams: RIPT-native lossy mode, 14-domain profile tuning, multiband BSQ/BIL/BIP, NoData mask integration, plus BVC lossless point clouds & splats below incumbents' lossy (maximum architectural lift)

All multipliers below are DIRECTIONAL and PENDING DESIGN-PARTNER VALIDATION. Tier 3 (RIPT / BVC native-format adoption) is where the architectural lift compounds; Tiers 0-2 are interim values that grow with the number and types of imagery formats adopted.

Shared properties

Two structural choices the C upstreams cannot retrofit.

Memory safety

Every Track B consumer library (turbo*-safe) is zero-unsafe. TurboWebP replaces the libwebp C library implicated in CVE-2023-4863 (heap buffer overflow exploited in the wild, September 2023). Every other drop-in replaces a C codec with the same class of risk.

Cross-architecture determinism

Byte-identical encoder output across AVX2 / NEON / WASM SIMD128 for TurboWebP, TurboJXL, TurboZstd, TurboLERC, with SVE2 and AVX-512 planned for future releases. Same input + same level + any host = same output bytes. The C upstreams do not offer this.

Talk to us about the suite.

Drop-in pilots run 2-6 weeks. Bring your own corpus and we will benchmark on it, so the numbers you decide on are measured against your data rather than ours.