Image Tutorial

Getting Started: Converting Image Files

Learn how to convert, edit, and AI-process images with Media Manipulator's image converter. This tutorial walks through every section of the image conversion panel — format selection, cropping, filters, text overlays, metadata management, and the AI image tools (Face Privacy, Remove Background, AI Upscale, Redact Text, and Remove Object).

1. Upload your image

Drag an image file into the upload zone, or click Choose File. The image converter accepts JPG, PNG, WebP, GIF, and many other formats. Files are deleted within 24 hours.

Loading the tool…

2. Pick the output format and quality

  • JPG — best for photos. Lossy compression with a quality slider.
  • PNG — lossless, supports transparency. Best for screenshots, logos, and graphics.
  • WebP — modern format with both lossy and lossless modes; smaller than JPG at similar quality.
  • GIF — palette-based, supports basic animation, useful for legacy use cases.

Quality (%) applies to JPG and WebP. 85 is a great default; lower for thumbnails, higher for hero images.

3. Resize, crop, filter, tint

Use Width and Height to resize. Leave either blank to scale proportionally. ImageMagick auto-orients the input first so EXIF-oriented JPEGs don't ship rotated to PNG/WebP.

Click Crop Image for a visual crop selector. The Filter dropdown applies a built-in effect: grayscale, sepia, blur, sharpen, swirl, barrel distortion, oil painting, vintage, emboss, charcoal, sketch, or a fixed rotation (45° / 90° / 180° / 270°). Tint Color lays a 30% colored tint over the image (skip by leaving black/#000000).

4. Text overlay

Add a watermark or caption directly to the image. Type your text into Overlay Text, then choose:

  • Text Size — 8–512 px.
  • Position — gravity-based anchor (corners, edges, center).
  • Text Color / Stroke Color — hex colors via the color pickers.
  • Stroke Width — 0 to disable the outline.
  • X Offset / Y Offset — pixel-level nudge from the gravity anchor.

5. Metadata

Choose how Media Manipulator handles EXIF, IPTC, XMP, and GPS metadata:

  • Keep all metadata — preserved from the original via exiftool.
  • Strip all metadata — removes camera, GPS, software, and timestamps. Recommended before sharing photos online.
  • Custom metadata — strip first, then rewrite the fields you supply. Includes:
    • Title, Artist/Author, Description, Copyright, User Comment, Keywords.
    • GPS removal/replacement (with optional precision rounding) and direction/timestamp/altitude/destination toggles.
    • EXIF/TIFF fields, GPS fields, and advanced IPTC/XMP/MakerNotes tags via an allowlist.

6. AI Image Tools (recommended)

The AI Image Tools panel runs one local GPU operation per conversion job. When you pick an operation, the standard ImageMagick pipeline (filters, tint, overlay, etc.) is skipped for that job so the AI tool has a clean input/output path.

  • Face Blur / Pixelate / Black Box — detects every face in the photo and obscures it with your chosen style. Great for sharing event photos without leaking attendee identities.
  • Remove Background — runs rembg with BiRefNet (general or portrait), ISNet, or U2Net. Output is forced to PNG so transparency is preserved. BiRefNet General is the recommended default for most photos.
  • AI Upscale — Real-ESRGAN ncnn Vulkan upscaler. Pick 4x (safest) or 2x (faster), and pick a model: realesrgan-x4plus for general photos, realesrgan-x4plus-anime for animation/illustration, or realesr-animevideov3 for anime video frames. PNG output by default.
  • Redact Text / PII — OCR-based redaction. Choose PII only (emails, phone numbers, SSNs, addresses, etc.) or All Text, then pick a redaction style (Black Box, Blur, or Pixelate). Best-effort — review the output before sharing.
  • Remove Object (LaMa Inpainting) — click and drag on the image preview to draw a rectangle over whatever you want to erase: a passerby in the background, a sign, a watermark, a pole. Drag the rectangle body to move it, drag the corner/edge handles to resize, and add as many rectangles as you need with another click-drag on empty space. When you submit, the server rasterizes a mask from your rectangles and runs simple-lama-inpainting on a dedicated GPU to reconstruct the area behind them. The output keeps your chosen format (JPG, PNG, WebP, GIF). Tips:
    • Draw the rectangle snugly around the object — large rectangles ask LaMa to hallucinate more background, which can look worse.
    • Several small rectangles usually beat one huge one, especially for thin objects.
    • LaMa works best when the surrounding pixels give the model enough context (sky, grass, plain walls, even-toned surfaces).

7. Convert and download

Click Convert File. A progress bar shows real-time job status. Once finished, the result auto-previews with an Original / Final toggle so you can compare. Use Download to save it, or revisit it later from the in-session Conversion History list.

Tips for great image output

  • For social uploads, use WebP at 80–85% quality for great-looking small files.
  • Pair Strip metadata with sharing photos online to avoid leaking GPS or device info.
  • Remove Background outputs are always PNG — pair them with a colored background in your designer of choice or stack them in a slide deck.
  • For AI Upscale, prefer 4x with realesrgan-x4plus on photos; 2x is a power-user option that can occasionally show framing artifacts on small inputs.
  • Run Redact Text before posting a screenshot of an email, ticket, or invoice.
  • Use Remove Object for one-off cleanup like a tourist in a vacation photo or a stray sign over a landscape — but expect to iterate on the rectangle size to find the cleanest result.
Try the Image Converter← Back to TutorialsHow it Works