verda-watermark 0.3.0

Python SDK for the Verda Enterprise Watermarking API

Verda Python SDK

Python SDK for the Verda Watermarking API.

Embed imperceptible watermarks into images, videos, and audio. Decode watermarks to trace content back to its source.

Install

pip install verda-watermark

Quick Start

from verda import VerdaClient

client = VerdaClient(api_key="vk_...")

# Encode a watermark
result = client.encode(file="photo.jpg")
print(result.watermark_ref)  # Unique reference for this watermark
print(result.download_url)   # Download the watermarked file

# Encode from a URL
result = client.encode(url="https://cdn.example.com/video.mp4")

# Decode / verify a watermark
result = client.decode(file="suspect.jpg")
print(result.match)          # True if watermark found
print(result.provenance)     # Origin info (owner, timestamp, etc.)

# Check credits
balance = client.credits()
print(f"${balance.balance_dollars:.2f} remaining")

Configuration

client = VerdaClient(
    api_key="vk_...",                                    # or set VERDA_API_KEY env var
    base_url="https://api.verda.ai/api/v2/enterprise",   # default
    poll_interval=5,                                      # seconds between job polls
    poll_timeout=300,                                     # max seconds to wait
)

Async Jobs

By default, encode() and decode() wait for the job to complete. For long-running jobs (video), use wait=False:

# Start encode without waiting
job_id = client.encode("long_video.mp4", wait=False)

# Check status later
job = client.get_job(job_id)
print(job.status)  # QUEUED, PROCESSING, COMPLETED, or FAILED

Job Management

# List your recent jobs
job_list = client.list_jobs(page=1, limit=20)
for job in job_list.jobs:
    print(f"{job.job_id}: {job.job_type} - {job.status}")

# Delete server-side files for a completed job
client.delete_job_files(job_id)

Watermark Registry (On-Prem)

For on-prem deployments where you run the codec yourself, use the registry to allocate watermark IDs centrally:

# Register a watermark before encoding locally
reg = client.register_watermark(
    content_type="audio",
    file_hash="sha256_of_original_file",
    file_format="wav",
    duration_seconds=180.0,
)
uid = reg["uid"]               # Embed this value with your local codec
ref = reg["watermark_ref"]     # Store this for your records

# After decoding, look up the watermark
entry = client.lookup_watermark(uid=uid)
if entry:
    print(f"Registered by: {entry.client_id}")
    print(f"Content type: {entry.content_type}")

Model Manifest (On-Prem)

manifest = client.get_model_manifest()
print(f"Latest encode version: {manifest.latest_encode_version}")
for v in manifest.versions:
    print(f"  v{v.version} (default={v.is_default_encode})")

On-Prem Mode

Point the client to your local server:

client = VerdaClient(api_key="vk_...", on_prem="http://localhost:8001")

# File-based
result = client.encode(file="speech.wav")

# Tensor-based (requires numpy + soundfile)
result = client.encode_tensor(audio_16k=waveform, sample_rate=16000)

Supported Formats

Type	Formats
Image	jpg, png, webp, bmp, gif, tiff, heic
Video	mp4, mov, avi, mkv, webm
Audio	mp3, wav, flac, aac, ogg, m4a

Pricing

New accounts receive a one-time free credit. See enterprise.verda.ai/api-docs for current pricing.

Error Handling

from verda import VerdaClient, VerdaError, InsufficientCreditsError

try:
    result = client.encode(file="photo.jpg")
except InsufficientCreditsError:
    print("Top up your credits at enterprise.verda.ai")
except VerdaError as e:
    print(f"Error: {e.message} (HTTP {e.status_code})")

License

MIT