From 8e45898772c3576ba26d90a54e5cc7391a4657c0 Mon Sep 17 00:00:00 2001
From: Benjamin Sparks <b.sparks@alugha.com>
Date: Tue, 18 Feb 2025 19:56:36 +0100
Subject: [PATCH] python3Packages.whisperx: init at 3.3.1

---
 pkgs/by-name/wh/whisperx/package.nix          |  1 +
 .../python-modules/whisperx/default.nix       | 87 +++++++++++++++++++
 pkgs/top-level/python-packages.nix            |  2 +
 3 files changed, 90 insertions(+)
 create mode 100644 pkgs/by-name/wh/whisperx/package.nix
 create mode 100644 pkgs/development/python-modules/whisperx/default.nix

diff --git a/pkgs/by-name/wh/whisperx/package.nix b/pkgs/by-name/wh/whisperx/package.nix
new file mode 100644
index 0000000000000..c5b60fb97af61
--- /dev/null
+++ b/pkgs/by-name/wh/whisperx/package.nix
@@ -0,0 +1 @@
+{ python3Packages }: with python3Packages; toPythonApplication whisperx
diff --git a/pkgs/development/python-modules/whisperx/default.nix b/pkgs/development/python-modules/whisperx/default.nix
new file mode 100644
index 0000000000000..91298a1b86c94
--- /dev/null
+++ b/pkgs/development/python-modules/whisperx/default.nix
@@ -0,0 +1,87 @@
+{
+  lib,
+  buildPythonPackage,
+  fetchFromGitHub,
+
+  setuptools,
+
+  ctranslate2,
+  ctranslate2-cpp,
+  faster-whisper,
+  ffmpeg,
+  nltk,
+  pandas,
+  pyannote-audio,
+  torch,
+  torchaudio,
+  transformers,
+
+  cudaSupport ? torch.cudaSupport,
+  cudaPackages,
+}:
+
+let
+  # inherit (cudaPackages) cudaFlags cudnn nccl;
+
+  ctranslate = ctranslate2.override {
+    ctranslate2-cpp = ctranslate2-cpp.override {
+      withCUDA = cudaSupport ? false;
+      withCuDNN = cudaSupport ? false;
+    };
+  };
+in
+buildPythonPackage rec {
+  pname = "whisperx";
+  version = "3.3.1";
+  pyproject = true;
+
+  src = fetchFromGitHub {
+    owner = "m-bain";
+    repo = "whisperX";
+    tag = "v${version}";
+    hash = "sha256-Wb9jKTs0rTdgjvcnKl1P8Uh6b5jzlkKgQBcJYA+7+W4=";
+  };
+
+  nativeBuildInputs = lib.optionals cudaSupport ( with cudaPackages; [
+
+  ]);
+
+  build-system = [ setuptools ];
+
+  dependencies = [
+    ctranslate
+    faster-whisper
+    nltk
+    pandas
+    pyannote-audio
+    torch
+    torchaudio
+    transformers
+  ];
+
+  pythonRelaxDeps = [
+    "ctranslate2"
+    "faster-whisper"
+  ];
+
+  pythonImportsCheck = [ "whisperx" ];
+
+  # No tests in repository
+  doCheck = false;
+
+  makeWrapperArgs = [ "--prefix PATH : ${lib.makeBinPath [ ffmpeg ]}" ];
+
+  meta = {
+    mainProgram = "whisperx";
+    description = "Automatic Speech Recognition with Word-level Timestamps (& Diarization)";
+    homepage = "https://github.com/m-bain/whisperX";
+    license = lib.licenses.bsd2;
+    maintainers = with lib.maintainers; [ bengsparks ];
+
+    platforms =
+      if cudaSupport then
+        lib.platforms.linux
+      else
+        lib.platforms.unix;
+  };
+}
diff --git a/pkgs/top-level/python-packages.nix b/pkgs/top-level/python-packages.nix
index 2c30a1d273f97..2c0e9d0101785 100644
--- a/pkgs/top-level/python-packages.nix
+++ b/pkgs/top-level/python-packages.nix
@@ -18324,6 +18324,8 @@ self: super: with self; {
 
   whispers = callPackage ../development/python-modules/whispers { };
 
+  whisperx = callPackage ../development/python-modules/whisperx { };
+
   whitenoise = callPackage ../development/python-modules/whitenoise { };
 
   whodap = callPackage ../development/python-modules/whodap { };