Multi-Modal Translation¶

from pauhu import Pauhu

client = Pauhu()

# Translate text in image
result = client.translate_image(
    image_path="sign.jpg",
    target="fi"
)

print(result.extracted_text)  # "Exit"
print(result.translation)      # "Uloskäynti"
print(result.confidence)       # 0.98

# Create translated image with text overlay
result = client.translate_image(
    image_path="infographic.png",
    target="fi",
    output="overlay"  # Replace text in image
)

result.save("infographic_fi.png")

# Process folder of images
results = client.translate_images(
    folder_path="/images/english",
    target="fi",
    output_folder="/images/finnish",
    formats=["png", "jpg", "tiff"]
)

from pauhu import Pauhu

client = Pauhu()

# Translate audio file
result = client.translate_audio(
    audio_path="meeting.mp3",
    source="en",
    target="fi",
    output="audio"  # Generate Finnish audio
)

result.save("meeting_fi.mp3")

# Access intermediate results
print(result.transcript)      # Original English text
print(result.translation)     # Finnish text
print(result.duration_ms)     # Audio duration

from pauhu.realtime import AudioStream

client = Pauhu()

async def translate_live():
    audio = AudioStream(device="microphone")

    async for segment in audio.transcribe_and_translate(
        source="en",
        target="fi",
        output="audio"
    ):
        # Play translated audio in real-time
        segment.play()

        # Or access text
        print(f"EN: {segment.source_text}")
        print(f"FI: {segment.target_text}")

# Generate Finnish speech from English text
result = client.text_to_speech(
    text="Welcome to Finland",
    source="en",
    target="fi",
    voice="female_1"  # Finnish voice
)

result.save("welcome_fi.mp3")

from pauhu import Pauhu

client = Pauhu()

# Generate translated subtitles
result = client.translate_video(
    video_path="presentation.mp4",
    target="fi",
    output="subtitles"
)

# Save SRT file
result.save_subtitles("presentation_fi.srt")

# Embed subtitles
result.embed_subtitles(
    output_path="presentation_subtitled.mp4"
)

# Full audio dubbing
result = client.translate_video(
    video_path="training.mp4",
    target="fi",
    output="dubbed",
    preserve_music=True,  # Keep background music
    voice_cloning=True    # Match original speaker
)

result.save("training_fi.mp4")

# Generate subtitles for multiple languages
targets = ["fi", "sv", "de", "fr"]

for lang in targets:
    result = client.translate_video(
        video_path="product_demo.mp4",
        target=lang,
        output="subtitles"
    )
    result.save_subtitles(f"product_demo_{lang}.srt")

# Use specific voice
result = client.translate_audio(
    audio_path="speech.mp3",
    target="fi",
    voice="fi_female_1",
    speed=1.0,           # Normal speed
    pitch=1.0            # Normal pitch
)

# Clone speaker's voice in target language
result = client.translate_audio(
    audio_path="ceo_speech.mp3",
    target="fi",
    voice_cloning=True,
    reference_audio="ceo_sample.mp3"  # 30+ seconds
)

from pauhu import Pauhu

client = Pauhu()

# High-quality settings for production
result = client.translate_audio(
    audio_path="audiobook.mp3",
    target="fi",
    quality={
        "transcription": "whisper-large",
        "translation": "quality",
        "tts": "neural-hd",
        "sample_rate": 48000,
        "bitrate": 320
    }
)

# Fast settings for real-time
result = client.translate_audio(
    audio_path="call.mp3",
    target="fi",
    quality={
        "transcription": "whisper-small",
        "translation": "fast",
        "tts": "neural",
        "sample_rate": 16000,
        "bitrate": 128
    }
)

from pauhu import Pauhu

client = Pauhu()

# Translate image
image_result = client.translate_image(
    image_path="sign.jpg",
    target="fi"
)
print(image_result.translation)

# Translate audio
audio_result = client.translate_audio(
    audio_path="speech.mp3",
    target="fi"
)
audio_result.save("speech_fi.mp3")

# Generate subtitles
video_result = client.translate_video(
    video_path="video.mp4",
    target="fi",
    output="subtitles"
)
video_result.save_subtitles("video_fi.srt")