octra-vanity

This is a gpu vanity address miner for Octra network, it brute-forces ed25519 keypairs until your address looks the way you want. You have to set the GPU parameter in the .bat file.

oct XXXXX xKp3mNvQr7wZsYdF2jBgTcLuA8eX4nRo
    ^^^^^
    you pick this part

---

requirements

• NVIDIA GPU (Maxwell or newer)
• CUDA Toolkit 11+

GPU	ARCH
RTX 4000 series	sm_89
RTX 3000 series	sm_86
RTX 2000 / GTX 1600	sm_75
GTX 1000 series	sm_61

---

build

Windows

also needs Visual Studio with Desktop development with C++ (free Community edition works).

cd vanity
build.bat

if it says cl.exe not found → open Visual Studio Installer → Modify → check Desktop development with C++ → install.

Linux

cd vanity
nvcc -O3 -arch=sm_89 -lm octra_vanity.cu -o octra_vanity
./octra_vanity

replace sm_89 with your GPU's compute capability (check with nvidia-smi --query-gpu=compute_cap --format=csv).

needs nvcc in PATH. on Ubuntu/Debian:

sudo apt install nvidia-cuda-toolkit

or install the full CUDA toolkit from nvidia for the latest version.

macOS

macOS dropped NVIDIA support in 2019 :(

---

usage

just run it, no flags needed:

octra_vanity.exe

picks the mode interactively:

1. Prefix      octHELLO...
2. Suffix      ...WORLD
3. Anywhere    ...DEGEN...
4. Repeating   octXXXXX... or ...ZZZZZ or 6 same chars anywhere

type your pattern, hit enter a few times for defaults, it starts mining.

---

how long will it take

roughly 25–35M keys/sec on an RTX 4090 (up from 3-5M after major optimizations!).

what you want	expected time
3-char prefix (octABC)	< 1 sec
4-char prefix (octABCD)	< 1 sec
5-char prefix (octHELLO)	~10-15 sec
6-char prefix	~10-15 min
7-char prefix	~10 hours
5 repeating (octXXXXX)	< 1 sec
6 repeating	~15 sec
7 repeating	~15 min

difficulty doubles roughly every extra character. plan accordingly.

---

optimizations

this miner is heavily optimized for maximum throughput:

1. Precomputed Tables for Ed25519 (Windowed Scalar Multiplication): Bypasses the slow "double-and-add" method for public key derivation by precomputing a table of curve points for 8-bit windows during initialization, drastically reducing the number of heavy GF math operations required per key.
2. Truncated SHA-512: Seed hash needs a SHA-512 digest, but Ed25519 only cares about the first 32 bytes. The SHA-512 implementation was rewritten to stop processing early and output directly into 32-bit integers, bypassing unnecessary memory writes.
3. 32-bit Base58 Encoding & Matching: The Base58 routine and pattern matchers work on arrays of 32-bit integers instead of 8-bit bytes, reducing division/modulo operations on the GPU by ~75%.
4. Dedicated Point Doubling: Replaced a generic unified ed_add with a highly specialized and simplified ed_double function for the elliptic curve math, saving temporary variables and field multiplications.
5. Register Pressure Reduction: Pack values efficiently to drop the register count per thread significantly (down to 188 registers per thread), allowing much better occupancy and switching speed on the GPU streaming multiprocessors.

---

output

when found, prints this:

  ╔══════════════════════════════════════════╗
  ║                 FOUND!                   ║
  ╚══════════════════════════════════════════╝

  Address     octHELLOxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
  Private key <base64>
  Tried       11,534,336 keys in 2s  (5,767,168 keys/s)

  Saved       wallet_vanity_HELLOxxx.json

the .json file is ready to drop into the webcli wallet importer as-is.

---

valid characters

base58 — no 0, O, I, l. everything else is fair game.

123456789ABCDEFGHJKLMNPQRSTUVWXYZabcdefghijkmnopqrstuvwxyz

---

notes

• each run uses a fresh random seed so two runs never duplicate
• the found wallet json matches the exact format the octra webcli expects
• Ctrl+C to cancel at any point