Batch Image Upscaling: How to Upscale 100+ Photos at Once With AI.

The Two Approaches: ZIP Batch vs. API

There are two ways to batch-upscale images on UpRes, and choosing between them depends on your technical comfort level and volume:

• ZIP batch upload (web app): Upload a ZIP file containing up to 50 images, select a model, wait for processing, download a ZIP of results. No code required. Works on Pro and Business plans.
• REST API (Business plan): POST each image URL to /v1/jobs, collect job IDs, poll for completion in parallel. Handles unlimited volume, no browser required, integrates into your own pipeline.

Method 1: ZIP Batch Upload (No Code)

The simplest path for photographers and designers who don't want to write code:

1. Compress your images into a ZIP file on your computer. Keep them under 50 images per ZIP and under 500 MB total (individual image limit is 50 MB each).
2. Open the UpRes web app and drag the ZIP onto the upload zone. The app detects ZIP files automatically and queues each image as a separate job.
3. Select your model. For product photos at consistent quality, Real-ESRGAN is fastest and most cost-effective. For images with faces, switch to Bria Increase Resolution.
4. Wait for the batch to complete — the job list updates in real time. Download the result ZIP when all jobs show "completed."

Time estimate: 50 product JPEGs (3–5 MB each) through Real-ESRGAN typically completes in 4–8 minutes on a Pro plan.

Method 2: REST API (Any Volume, Programmatic)

For volumes over 50 images, recurring batch jobs, or integration into a product, the API is the correct approach. UpRes Business ($49/month) includes full API access.

The basic loop in Python:

import requests
import time

API_KEY = "your_api_key_here"
IMAGE_URLS = ["https://your-cdn.com/img001.jpg", ...]  # list of 100+ URLs
MODEL = "flare"

# Submit all jobs in parallel
jobs = []
for url in IMAGE_URLS:
    r = requests.post(
        "https://api.upres.ai/v1/jobs",
        json={"image_url": url, "model": MODEL, "scale": 4},
        headers={"Authorization": f"Bearer {API_KEY}"},
    )
    jobs.append(r.json()["id"])

# Poll until all complete
results = {}
pending = set(jobs)
while pending:
    for job_id in list(pending):
        r = requests.get(
            f"https://api.upres.ai/v1/jobs/{job_id}",
            headers={"Authorization": f"Bearer {API_KEY}"},
        )
        job = r.json()
        if job["status"] == "completed":
            results[job_id] = job["result_url"]
            pending.discard(job_id)
    if pending:
        time.sleep(5)

print(f"Done — {len(results)} images processed")

Full API reference at /docs/api. The OpenAPI 3.1 spec is at https://api.upres.ai/v1/openapi.json if you want to generate a typed client in any language.

Choosing the Right Model for Batch Jobs

At volume, model choice has a large impact on cost and speed. Here's the decision matrix for the most common batch scenarios:

For a 350-image product batch through Real-ESRGAN, PAYG needs 350 credits. On the Pro plan's 100 included images, the remaining 250 images need 250 credits. See the batch upscaling guide for a full cost breakdown.

Common Batch Workflow Mistakes

• Using the wrong scale factor: a 3200×3200 output from an 800×800 input means 4× upscaling. For a 3200×3200 target, request 4x. If you specify 2x and then try to scale again, you get two generations of artifacts.
• Pre-compressing inputs: Don't ZIP already-compressed JPEGs with aggressive JPEG compression again. Each JPEG recompression cycle degrades quality. Use the original files as inputs.
• Not rate-limiting API calls: The API accepts 60 requests/minute. Submit in batches of 50–55 with a 1-second sleep between batches to avoid 429 errors.
• Single model for all content: A batch of mixed portraits and landscapes should use Bria for the portraits and Real-ESRGAN for the landscapes — not one model for everything.

Integration Patterns

Beyond one-off batch jobs, teams often integrate upscaling into larger pipelines:

• E-commerce image pipeline: Ingest supplier images → check resolution → if under threshold, POST to UpRes API → replace with upscaled version before publishing to storefront.
• DAM (Digital Asset Management) integration:On asset upload to your DAM, trigger a webhook that sends the file to UpRes and stores the upscaled version alongside the original.
• Print-on-demand quality check: Before sending to a print vendor, run all uploaded images through Real-ESRGAN if they're under 300 DPI at the target print size.

The API integration guide covers webhook patterns and error handling in detail.

FAQ