SVG and text rendering

Rendering text is a common first step for labels, captions, and watermark building. libvips renders text through Pango, so you get markup (<i> italic, <span foreground=...> colour), word wrapping, and alignment. This recipe renders three text styles — basic, wrapped, and Pango markup — each composited onto a transparent 640 × 120 RGBA canvas and exported as PNG with alpha preserved.

Input

The input is an empty transparent SVG canvas of known dimensions; it acts as a neutral base and is not composited in this recipe.

Empty transparent SVG canvas used as the text-rendering base — **Input: SVG**800 × 200, transparent

The reference output was generated with govips v2.19.0 / libvips 8.17.3. Cross-binding pixel comparison is still pending.

Basic single-line text

One line of text at DejaVu Sans 28, rendered to RGBA and placed at the top-left of the canvas.

Single-line VIPS Handbook text on a transparent canvas — **text-basic: PNG**DejaVu Sans 28, RGBA

txt, err := vips.Text(&vips.TextParams{
    Text: "VIPS Handbook",
    Font: "DejaVu Sans 28",
    DPI:  72,
    RGBA: true,
    Wrap: vips.TextWrapWord,
})

await sharp({ create: { width: 640, height: 120, channels: 4,
                         background: { r: 0, g: 0, b: 0, alpha: 0 } } })
  .composite([{ input: { text: { text: 'VIPS Handbook',
                                  font: 'DejaVu Sans 28', rgba: true } },
                left: 24, top: 24, blend: 'over' }])
  .png()
  .toFile('text-basic.png');

canvas = pyvips.Image.black(640, 120, bands=4).cast("uchar").copy(interpretation="srgb")
text = pyvips.Image.text("VIPS Handbook", font="DejaVu Sans 28", rgba=True)
result = canvas.composite2(text, "over", x=24, y=24)
result.write_to_file("text-basic.png")

vips black canvas.png 640 120 --bands 4
vips copy canvas.png rgba.png --interpretation srgb
vips text txt.png "VIPS Handbook" --font 'DejaVu Sans 28' --rgba
vips composite2 rgba.png txt.png 'text-basic.png' over --x 24 --y 24

Wrapped text

A longer sentence wrapped to a 320 pixel width at DejaVu Sans 20. The transparent canvas keeps the output the same 640 × 120 size across bindings.

A long sentence wrapped to a 320-pixel width inside the transparent canvas — **text-wrapped: PNG**DejaVu Sans 20, width 320, word wrap

txt, err := vips.Text(&vips.TextParams{
    Text: "The libvips demand-driven pipeline keeps memory flat while processing large images in production.",
    Font: "DejaVu Sans 20",
    DPI:  72,
    Width: 320,
    RGBA: true,
    Wrap: vips.TextWrapWord,
})

await sharp({ create: { width: 640, height: 120, channels: 4,
                         background: { r: 0, g: 0, b: 0, alpha: 0 } } })
  .composite([{ input: { text: {
      text: 'The libvips demand-driven pipeline keeps memory flat while processing large images in production.',
      font: 'DejaVu Sans 20', width: 320, rgba: true } },
      left: 24, top: 24, blend: 'over' }])
  .png()
  .toFile('text-wrapped.png');

canvas = pyvips.Image.black(640, 120, bands=4).cast("uchar").copy(interpretation="srgb")
text = pyvips.Image.text(
    "The libvips demand-driven pipeline keeps memory flat while processing large images in production.",
    font="DejaVu Sans 20", width=320, rgba=True)
result = canvas.composite2(text, "over", x=24, y=24)
result.write_to_file("text-wrapped.png")

vips black canvas.png 640 120 --bands 4
vips copy canvas.png rgba.png --interpretation srgb
vips text txt.png "The libvips demand-driven pipeline keeps memory flat while processing large images in production." --font 'DejaVu Sans 20' --rgba --width 320
vips composite2 rgba.png txt.png 'text-wrapped.png' over --x 24 --y 24

Pango markup

Pango markup mixes font weight, colour, and italic in a single string: bold violet VIPS followed by an italic Handbook.

txt, err := vips.Text(&vips.TextParams{
    Text: "<span font_desc='DejaVu Sans' font_size='30000' font_weight='bold' foreground='#5d2f86'>VIPS</span> <i>Handbook</i>",
    Font: "DejaVu Sans",
    DPI:  72,
    RGBA: true,
    Wrap: vips.TextWrapWord,
})

await sharp({ create: { width: 640, height: 120, channels: 4,
                         background: { r: 0, g: 0, b: 0, alpha: 0 } } })
  .composite([{ input: { text: {
      text: "<span font_desc='DejaVu Sans' font_size='30000' font_weight='bold' foreground='#5d2f86'>VIPS</span> <i>Handbook</i>",
      rgba: true } },
      left: 24, top: 24, blend: 'over' }])
  .png()
  .toFile('text-pango.png');

canvas = pyvips.Image.black(640, 120, bands=4).cast("uchar").copy(interpretation="srgb")
text = pyvips.Image.text(
    "<span font_desc='DejaVu Sans' font_size='30000' font_weight='bold' foreground='#5d2f86'>VIPS</span> <i>Handbook</i>",
    rgba=True)
result = canvas.composite2(text, "over", x=24, y=24)
result.write_to_file("text-pango.png")

vips black canvas.png 640 120 --bands 4
vips copy canvas.png rgba.png --interpretation srgb
vips text txt.png "<span font_desc='DejaVu Sans' font_size='30000' font_weight='bold' foreground='#5d2f86'>VIPS</span> <i>Handbook</i>" --rgba
vips composite2 rgba.png txt.png 'text-pango.png' over --x 24 --y 24

How each binding exposes text rendering

Text rendering is the one “drawing”-like operation with a shared portable core, but the API shape differs:

govips has a hand-written vips.Text(*TextParams) helper that returns a standalone *ImageRef.
sharp has no standalone text method; text is rendered as an overlay via composite([{ input: { text: {...} } }]).
pyvips exposes pyvips.Image.text(...) directly.
vips CLI exposes vips text.

The wrapped and Pango variants here were intentionally composited onto a fixed-size transparent canvas so the output dimensions are identical across all four bindings (640 × 120). If you render text at its natural size instead, the glyph metrics depend on the installed font set and DPI, and the output can differ between bindings.