FEAT: New Audio Convertors

doc/api.rst

@@ -406,10 +406,14 @@ API Reference
     AddTextImageConverter
     AnsiAttackConverter
     AsciiArtConverter
+    AudioEchoConverter
     AsciiSmugglerConverter
     AskToDecodeConverter
     AtbashConverter
     AudioFrequencyConverter
+    AudioSpeedConverter
+    AudioVolumeConverter
+    AudioWhiteNoiseConverter
     AzureSpeechAudioToTextConverter
     AzureSpeechTextToAudioConverter
     Base2048Converter
@@ -441,6 +445,7 @@ API Reference
     MaliciousQuestionGeneratorConverter
     MathObfuscationConverter
     MathPromptConverter
+    MultiLanguageTranslationConverter
     MorseConverter
     NatoConverter
     NegationTrapConverter

doc/code/converters/1_text_to_text_converters.py

@@ -245,6 +245,7 @@
     DenylistConverter,
     MaliciousQuestionGeneratorConverter,
     MathPromptConverter,
+    MultiLanguageTranslationConverter,
     NoiseConverter,
     PersuasionConverter,
     RandomTranslationConverter,
@@ -279,12 +280,20 @@
 translation_converter = TranslationConverter(converter_target=attack_llm, language="French")
 print("Translation (French):", await translation_converter.convert_async(prompt=prompt))  # type: ignore
 
-# Random translation through multiple languages
+# Random translation translates each word to a random language
 random_translation_converter = RandomTranslationConverter(
     converter_target=attack_llm, languages=["French", "German", "Spanish", "English"]
 )
 print("Random Translation:", await random_translation_converter.convert_async(prompt=prompt))  # type: ignore
 
+# Multi-language translation splits the prompt into segments and translates each to a different language.
+# Unlike RandomTranslationConverter (word-level, random), this operates on contiguous multi-word segments
+# with deterministic language assignment — segment i maps to languages[i].
+multi_lang_converter = MultiLanguageTranslationConverter(
+    converter_target=attack_llm, languages=["French", "Spanish", "Italian"]
+)
+print("Multi-Language Translation:", await multi_lang_converter.convert_async(prompt=prompt))  # type: ignore
+
 # Tense changes verb tense
 tense_converter = TenseConverter(converter_target=attack_llm, tense="far future")
 print("Tense (future):", await tense_converter.convert_async(prompt=prompt))  # type: ignore

doc/code/converters/2_audio_converters.ipynb

@@ -15,7 +15,7 @@
     "\n",
     "- **[Text to Audio](#text-to-audio)**: Convert text into spoken audio files\n",
     "- **[Audio to Text](#audio-to-text)**: Transcribe audio files into text\n",
-    "- **[Audio to Audio](#audio-to-audio)**: Modify audio files (e.g., frequency changes)"
+    "- **[Audio to Audio](#audio-to-audio)**: Modify audio files (speed, volume, echo, frequency, noise)"
    ]
   },
   {
@@ -39,16 +39,14 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Found default environment files: ['/home/vscode/.pyrit/.env', '/home/vscode/.pyrit/.env.local']\n",
-      "Loaded environment file: /home/vscode/.pyrit/.env\n",
-      "Loaded environment file: /home/vscode/.pyrit/.env.local\n"
+      "No default environment files found. Using system environment variables only.\n"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "audio_path: /workspace/dbdata/prompt-memory-entries/audio/1767054630224156.wav\n"
+      "audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385538448420.wav\n"
      ]
     }
    ],
@@ -122,7 +120,15 @@
     "<a id=\"audio-to-audio\"></a>\n",
     "## Audio to Audio\n",
     "\n",
-    "The `AudioFrequencyConverter` modifies audio files by increasing their frequency, enabling the probing of audio modality targets with heightened frequencies."
+    "Audio-to-audio converters modify existing audio files. All of these converters accept `audio_path` input\n",
+    "and produce `audio_path` output, preserving the original sample rate, bit depth, and channel count.\n",
+    "\n",
+    "Available converters:\n",
+    "- **`AudioFrequencyConverter`** — Shifts the audio frequency (pitch) higher\n",
+    "- **`AudioSpeedConverter`** — Changes playback speed without altering pitch\n",
+    "- **`AudioVolumeConverter`** — Scales the amplitude (louder or quieter)\n",
+    "- **`AudioEchoConverter`** — Adds an echo effect with configurable delay and decay\n",
+    "- **`AudioWhiteNoiseConverter`** — Mixes white noise into the signal"
    ]
   },
   {
@@ -135,21 +141,100 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "audio_path: /workspace/dbdata/prompt-memory-entries/audio/1767054635425663.wav\n"
+      "Frequency shift: audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540595766.wav\n",
+      "Speed (0.5x): audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540602793.wav\n",
+      "Volume (2x): audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540607791.wav\n",
+      "Echo: audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540611791.wav\n",
+      "White noise: audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540617281.wav\n"
      ]
     }
    ],
    "source": [
-    "from pyrit.prompt_converter import AudioFrequencyConverter\n",
+    "from pyrit.prompt_converter import (\n",
+    "    AudioEchoConverter,\n",
+    "    AudioFrequencyConverter,\n",
+    "    AudioSpeedConverter,\n",
+    "    AudioVolumeConverter,\n",
+    "    AudioWhiteNoiseConverter,\n",
+    ")\n",
     "\n",
     "# Use audio file created above\n",
     "assert os.path.exists(audio_convert_result.output_text)\n",
     "prompt = str(pathlib.Path(DB_DATA_PATH) / \"dbdata\" / \"audio\" / audio_convert_result.output_text)\n",
     "\n",
+    "# Frequency shift — increases the audio frequency (pitch)\n",
     "audio_frequency_converter = AudioFrequencyConverter()\n",
-    "converted_audio_file = await audio_frequency_converter.convert_async(prompt=prompt)  # type: ignore\n",
+    "converted = await audio_frequency_converter.convert_async(prompt=prompt)  # type: ignore\n",
+    "print(\"Frequency shift:\", converted)\n",
+    "\n",
+    "# Speed change — speeds up (>1.0) or slows down (<1.0) without pitch change\n",
+    "audio_speed_converter = AudioSpeedConverter(speed_factor=0.5)\n",
+    "converted = await audio_speed_converter.convert_async(prompt=prompt)  # type: ignore\n",
+    "print(\"Speed (0.5x):\", converted)\n",
+    "\n",
+    "# Volume scaling — amplifies (>1.0) or reduces (<1.0) the audio amplitude\n",
+    "audio_volume_converter = AudioVolumeConverter(volume_factor=2.0)\n",
+    "converted = await audio_volume_converter.convert_async(prompt=prompt)  # type: ignore\n",
+    "print(\"Volume (2x):\", converted)\n",
+    "\n",
+    "# Echo — adds a delayed, attenuated copy of the signal\n",
+    "audio_echo_converter = AudioEchoConverter(delay=0.3, decay=0.5)\n",
+    "converted = await audio_echo_converter.convert_async(prompt=prompt)  # type: ignore\n",
+    "print(\"Echo:\", converted)\n",
+    "\n",
+    "# White noise — mixes random noise into the audio\n",
+    "audio_noise_converter = AudioWhiteNoiseConverter(noise_scale=0.05)\n",
+    "converted = await audio_noise_converter.convert_async(prompt=prompt)  # type: ignore\n",
+    "print(\"White noise:\", converted)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "7",
+   "metadata": {},
+   "source": [
+    "### Chaining Audio Converters\n",
+    "\n",
+    "Audio-to-audio converters can be chained together to build a multi-step audio perturbation pipeline.\n",
+    "Each converter takes the output of the previous one as input."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "8",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "AudioSpeedConverter: audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540632000.wav\n",
+      "AudioVolumeConverter: audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540650996.wav\n",
+      "AudioEchoConverter: audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540668320.wav\n",
+      "AudioWhiteNoiseConverter: audio_path: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540685896.wav\n",
+      "\n",
+      "Final output: PyRIT\\dbdata\\prompt-memory-entries\\audio\\1771385540685896.wav\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Chain: slow down → increase volume → add echo → add white noise\n",
+    "pipeline = [\n",
+    "    AudioSpeedConverter(speed_factor=0.5),\n",
+    "    AudioVolumeConverter(volume_factor=1.5),\n",
+    "    AudioEchoConverter(delay=0.3, decay=0.5),\n",
+    "    AudioWhiteNoiseConverter(noise_scale=0.02),\n",
+    "]\n",
+    "\n",
+    "# Start with the original audio file\n",
+    "current_prompt = prompt\n",
+    "for converter in pipeline:\n",
+    "    result = await converter.convert_async(prompt=current_prompt)  # type: ignore\n",
+    "    current_prompt = result.output_text\n",
+    "    print(f\"{converter.__class__.__name__}: {result}\")\n",
     "\n",
-    "print(converted_audio_file)"
+    "print(f\"\\nFinal output: {current_prompt}\")"
    ]
   }
  ],

doc/code/converters/2_audio_converters.py

@@ -20,7 +20,7 @@
 #
 # - **[Text to Audio](#text-to-audio)**: Convert text into spoken audio files
 # - **[Audio to Text](#audio-to-text)**: Transcribe audio files into text
-# - **[Audio to Audio](#audio-to-audio)**: Modify audio files (e.g., frequency changes)
+# - **[Audio to Audio](#audio-to-audio)**: Modify audio files (speed, volume, echo, frequency, noise)
 
 # %% [markdown]
 # <a id="text-to-audio"></a>
@@ -73,16 +73,74 @@
 # <a id="audio-to-audio"></a>
 # ## Audio to Audio
 #
-# The `AudioFrequencyConverter` modifies audio files by increasing their frequency, enabling the probing of audio modality targets with heightened frequencies.
+# Audio-to-audio converters modify existing audio files. All of these converters accept `audio_path` input
+# and produce `audio_path` output, preserving the original sample rate, bit depth, and channel count.
+#
+# Available converters:
+# - **`AudioFrequencyConverter`** — Shifts the audio frequency (pitch) higher
+# - **`AudioSpeedConverter`** — Changes playback speed without altering pitch
+# - **`AudioVolumeConverter`** — Scales the amplitude (louder or quieter)
+# - **`AudioEchoConverter`** — Adds an echo effect with configurable delay and decay
+# - **`AudioWhiteNoiseConverter`** — Mixes white noise into the signal
 
 # %%
-from pyrit.prompt_converter import AudioFrequencyConverter
+from pyrit.prompt_converter import (
+    AudioEchoConverter,
+    AudioFrequencyConverter,
+    AudioSpeedConverter,
+    AudioVolumeConverter,
+    AudioWhiteNoiseConverter,
+)
 
 # Use audio file created above
 assert os.path.exists(audio_convert_result.output_text)
 prompt = str(pathlib.Path(DB_DATA_PATH) / "dbdata" / "audio" / audio_convert_result.output_text)
 
+# Frequency shift — increases the audio frequency (pitch)
 audio_frequency_converter = AudioFrequencyConverter()
-converted_audio_file = await audio_frequency_converter.convert_async(prompt=prompt)  # type: ignore
+converted = await audio_frequency_converter.convert_async(prompt=prompt)  # type: ignore
+print("Frequency shift:", converted)
+
+# Speed change — speeds up (>1.0) or slows down (<1.0) without pitch change
+audio_speed_converter = AudioSpeedConverter(speed_factor=0.5)
+converted = await audio_speed_converter.convert_async(prompt=prompt)  # type: ignore
+print("Speed (0.5x):", converted)
+
+# Volume scaling — amplifies (>1.0) or reduces (<1.0) the audio amplitude
+audio_volume_converter = AudioVolumeConverter(volume_factor=2.0)
+converted = await audio_volume_converter.convert_async(prompt=prompt)  # type: ignore
+print("Volume (2x):", converted)
+
+# Echo — adds a delayed, attenuated copy of the signal
+audio_echo_converter = AudioEchoConverter(delay=0.3, decay=0.5)
+converted = await audio_echo_converter.convert_async(prompt=prompt)  # type: ignore
+print("Echo:", converted)
+
+# White noise — mixes random noise into the audio
+audio_noise_converter = AudioWhiteNoiseConverter(noise_scale=0.05)
+converted = await audio_noise_converter.convert_async(prompt=prompt)  # type: ignore
+print("White noise:", converted)
 
-print(converted_audio_file)
+# %% [markdown]
+# ### Chaining Audio Converters
+#
+# Audio-to-audio converters can be chained together to build a multi-step audio perturbation pipeline.
+# Each converter takes the output of the previous one as input.
+
+# %%
+# Chain: slow down → increase volume → add echo → add white noise
+pipeline = [
+    AudioSpeedConverter(speed_factor=0.5),
+    AudioVolumeConverter(volume_factor=1.5),
+    AudioEchoConverter(delay=0.3, decay=0.5),
+    AudioWhiteNoiseConverter(noise_scale=0.02),
+]
+
+# Start with the original audio file
+current_prompt = prompt
+for converter in pipeline:
+    result = await converter.convert_async(prompt=current_prompt)  # type: ignore
+    current_prompt = result.output_text
+    print(f"{converter.__class__.__name__}: {result}")
+
+print(f"\nFinal output: {current_prompt}")

pyrit/prompt_converter/__init__.py

@@ -18,7 +18,11 @@
 from pyrit.prompt_converter.ascii_art_converter import AsciiArtConverter
 from pyrit.prompt_converter.ask_to_decode_converter import AskToDecodeConverter
 from pyrit.prompt_converter.atbash_converter import AtbashConverter
+from pyrit.prompt_converter.audio_echo_converter import AudioEchoConverter
 from pyrit.prompt_converter.audio_frequency_converter import AudioFrequencyConverter
+from pyrit.prompt_converter.audio_speed_converter import AudioSpeedConverter
+from pyrit.prompt_converter.audio_volume_converter import AudioVolumeConverter
+from pyrit.prompt_converter.audio_white_noise_converter import AudioWhiteNoiseConverter
 from pyrit.prompt_converter.azure_speech_audio_to_text_converter import AzureSpeechAudioToTextConverter
 from pyrit.prompt_converter.azure_speech_text_to_audio_converter import AzureSpeechTextToAudioConverter
 from pyrit.prompt_converter.base64_converter import Base64Converter
@@ -47,6 +51,7 @@
 from pyrit.prompt_converter.math_obfuscation_converter import MathObfuscationConverter
 from pyrit.prompt_converter.math_prompt_converter import MathPromptConverter
 from pyrit.prompt_converter.morse_converter import MorseConverter
+from pyrit.prompt_converter.multi_language_translation_converter import MultiLanguageTranslationConverter
 from pyrit.prompt_converter.nato_converter import NatoConverter
 from pyrit.prompt_converter.negation_trap_converter import NegationTrapConverter
 from pyrit.prompt_converter.noise_converter import NoiseConverter
@@ -110,7 +115,11 @@
     "AsciiSmugglerConverter",
     "AskToDecodeConverter",
     "AtbashConverter",
+    "AudioEchoConverter",
     "AudioFrequencyConverter",
+    "AudioSpeedConverter",
+    "AudioVolumeConverter",
+    "AudioWhiteNoiseConverter",
     "AzureSpeechAudioToTextConverter",
     "AzureSpeechTextToAudioConverter",
     "Base2048Converter",
@@ -142,6 +151,7 @@
     "MathObfuscationConverter",
     "MathPromptConverter",
     "MorseConverter",
+    "MultiLanguageTranslationConverter",
     "NatoConverter",
     "NegationTrapConverter",
     "NoiseConverter",