Short answer

draw = ImageDraw.Draw(img), load a font with ImageFont.truetype("path/to/font.ttf", size), then draw.text((x, y), "text", font=font, fill="white"). To centre text, pass the centre point and anchor="mm".

The basics

add_text.py
from PIL import Image, ImageDraw, ImageFont

img = Image.open("photo.jpg")
draw = ImageDraw.Draw(img)
font = ImageFont.truetype("DejaVuSans-Bold.ttf", 72)

draw.text((60, 60), "Hello, Pillow", font=font, fill="white")
img.save("hello.jpg", quality=90)
Tested on ✓ Linux ✓ Windows ✓ macOS
A launch pad sunset photo with the words Hello, Pillow in white in the top-left corner
Real output of the script above.

Three things to know from the start:

Positioning with anchors

The anchor argument says which point of the text sits at (x, y). It is two letters. The first is horizontal: l = left, m = middle, r = right. The second is vertical: a = ascender, t = top, m = middle, s = baseline, b = bottom, d = descender. The red dot is the (x, y) each call received:

A grid of nine anchor combinations showing where the text sits relative to a red dot
Real output of the script below.
anchors.py
from PIL import Image, ImageDraw, ImageFont

font = ImageFont.truetype("DejaVuSans.ttf", 30)
label = ImageFont.truetype("DejaVuSans.ttf", 16)
img = Image.new("RGB", (1050, 420), "white")
draw = ImageDraw.Draw(img)

for row, v in enumerate("ams"):            # ascender, middle, baseline
    for col, h in enumerate("lmr"):        # left, middle, right
        x, y = 175 + col * 350, 85 + row * 130
        anchor = h + v
        draw.line((x - 165, y, x + 165, y), fill="#cbd5e1")
        draw.line((x, y - 50, x, y + 50), fill="#cbd5e1")
        draw.text((x, y), "Typography", font=font, fill="#0f172a", anchor=anchor)
        draw.ellipse((x - 5, y - 5, x + 5, y + 5), fill="#dc2626")
        draw.text((x - 165, y - 62), f'anchor="{anchor}"', font=label, fill="#2563eb")
img.save("anchors.png")
Tested on ✓ Linux ✓ Windows ✓ macOS

Anchors are the cleanest way to align text: "mm" to centre, "rs" to right-align numbers on a baseline, "ls" for captions along the bottom. They replace the old pattern of measuring text with textsize(), which was removed in Pillow 10.

Centre text on an image

centre.py
from PIL import Image, ImageDraw, ImageFont

img = Image.open("photo.jpg")
draw = ImageDraw.Draw(img)
font = ImageFont.truetype("DejaVuSans-Bold.ttf", 110)

draw.text((img.width / 2, img.height * 0.3), "ARTEMIS II", font=font,
          fill="white", anchor="mm")
img.save("centred.jpg", quality=90)
Tested on ✓ Linux ✓ Windows ✓ macOS
The words ARTEMIS II centred horizontally over a launch pad photo
Real output (horizontally centred at 30% of the height, to stay clear of the towers).

"mm" centres the font's line box. For a single word without descenders that can look a few pixels high; the textsize page shows how to centre the visible ink exactly instead.

Make text readable: outline or background

Outline (stroke)

stroke_width and stroke_fill draw an outline around each glyph. It keeps text readable on any background:

outline.py
from PIL import Image, ImageDraw, ImageFont

img = Image.open("photo.jpg")
draw = ImageDraw.Draw(img)
font = ImageFont.truetype("DejaVuSans-Bold.ttf", 120)

draw.text((img.width / 2, img.height * 0.17), "LAUNCH DAY", font=font, anchor="mm",
          fill="white", stroke_width=6, stroke_fill="black")
img.save("outline.jpg", quality=90)
Tested on ✓ Linux ✓ Windows ✓ macOS
The words LAUNCH DAY in white letters with a black outline over a sunset sky
Real output.

Semi-transparent box behind the text

Drawing with an alpha value onto an RGB image does not blend. Draw on a transparent RGBA layer and composite it:

caption_box.py
from PIL import Image, ImageDraw, ImageFont

img = Image.open("photo.jpg").convert("RGBA")
overlay = Image.new("RGBA", img.size, (0, 0, 0, 0))     # transparent layer
draw = ImageDraw.Draw(overlay)
font = ImageFont.truetype("DejaVuSans.ttf", 56)

text = "Launch Pad 39B at sunset, 31 January 2026"
x, y = 60, 120                      # over the bright sky, where a backdrop is needed
left, top, right, bottom = draw.textbbox((x, y), text, font=font, anchor="ls")
pad = 24
draw.rounded_rectangle((left - pad, top - pad, right + pad, bottom + pad),
                       radius=18, fill=(0, 0, 0, 150))       # 150/255 opacity
draw.text((x, y), text, font=font, fill=(255, 255, 255, 255), anchor="ls")

result = Image.alpha_composite(img, overlay).convert("RGB")
result.save("caption.jpg", quality=90)
Tested on ✓ Linux ✓ Windows ✓ macOS
A caption in white text on a translucent dark rounded box over the sunset sky
Real output.

Wrap long text to a width

Pillow does not wrap text for you. Measure words with font.getlength() (the advance width in pixels) and build lines yourself. textwrap from the standard library counts characters, not pixels, so it wraps proportional fonts unevenly.

wrap_text.py
from PIL import Image, ImageDraw, ImageFont


def wrap(text, font, max_width):
    """Split text into lines that fit max_width pixels."""
    lines, line = [], ""
    for word in text.split():
        trial = f"{line} {word}".strip()
        if font.getlength(trial) <= max_width:
            line = trial
        else:
            if line:
                lines.append(line)
            line = word
    if line:
        lines.append(line)
    return "\n".join(lines)


quote = ("Pillow can open, edit and save dozens of image formats, "
         "and drawing text is one of the most common things people use it for.")
font = ImageFont.truetype("DejaVuSans.ttf", 40)
img = Image.new("RGB", (900, 420), "#0f172a")
draw = ImageDraw.Draw(img)

wrapped = wrap(quote, font, max_width=img.width - 160)
print(wrapped)
draw.multiline_text((img.width / 2, img.height / 2), wrapped, font=font, fill="white",
                    anchor="mm", align="center", spacing=14)
img.save("wrapped.png")
Tested on ✓ Linux ✓ Windows ✓ macOS
Output
Pillow can open, edit and save
dozens of image formats, and
drawing text is one of the most
common things people use it for.
A paragraph of white text wrapped into centred lines on a dark background
Real output.

Shrink text to fit a box

For titles of unknown length, such as product names or generated thumbnails, step the font size down until the text fits:

fit_text.py
from PIL import ImageDraw, ImageFont, Image


def fit_font(draw, text, font_path, box_w, box_h, start=200, minimum=8):
    """Largest font size at which text fits inside box_w x box_h."""
    size = start
    while size > minimum:
        font = ImageFont.truetype(font_path, size)
        left, top, right, bottom = draw.multiline_textbbox((0, 0), text, font=font)
        if right - left <= box_w and bottom - top <= box_h:
            return font
        size -= 2
    return ImageFont.truetype(font_path, minimum)


draw = ImageDraw.Draw(Image.new("RGB", (1, 1)))
for text in ["Hi", "Hello, Pillow", "A much longer headline"]:
    font = fit_font(draw, text, "DejaVuSans-Bold.ttf", box_w=600, box_h=120)
    print(f"{text!r:<26} -> size {font.size}")
Tested on ✓ Linux ✓ Windows ✓ macOS
Output
'Hi'                       -> size 158
'Hello, Pillow'            -> size 84
'A much longer headline'   -> size 44

For many sizes, a binary search is faster than stepping down by 2, but for one-off images this is plenty.

Rotated text and watermarks

draw.text() cannot rotate. Draw onto a separate transparent image, rotate that, and composite it:

rotated_text.py
from PIL import Image, ImageDraw, ImageFont

img = Image.open("photo.jpg").convert("RGBA")
font = ImageFont.truetype("DejaVuSans-Bold.ttf", 90)

# 1. Draw the text on its own transparent layer, just big enough for it.
left, top, right, bottom = font.getbbox("SAMPLE")
layer = Image.new("RGBA", (right + 20, bottom + 20), (0, 0, 0, 0))
ImageDraw.Draw(layer).text((10, 10), "SAMPLE", font=font, fill=(255, 255, 255, 140))

# 2. Rotate the layer (expand=True keeps the corners), then paste it using itself as the mask.
layer = layer.rotate(30, expand=True, resample=Image.BICUBIC)
img.alpha_composite(layer, ((img.width - layer.width) // 2, (img.height - layer.height) // 2))
img.convert("RGB").save("rotated.jpg", quality=90)
Tested on ✓ Linux ✓ Windows ✓ macOS
Semi-transparent white text SAMPLE rotated 30 degrees across the centre of a photo
Real output.

For tiled, logo or positioned watermarks, the watermark generator builds the code for you.

Non-Latin scripts and emoji

Arabic, Hebrew, Hindi and other complex scripts need the Raqm layout engine for correct shaping and right-to-left order. Pillow's Linux and macOS wheels include it; on Windows it also needs the FriBiDi DLL. Every script also needs a font that contains its glyphs:

scripts.py
from PIL import Image, ImageDraw, ImageFont, features

print("raqm layout engine available:", features.check("raqm"))

img = Image.new("RGB", (760, 330), "white")
draw = ImageDraw.Draw(img)
latin = ImageFont.truetype("DejaVuSans.ttf", 44)
cjk = ImageFont.truetype("/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc", 44)
emoji = ImageFont.truetype("/usr/share/fonts/truetype/noto/NotoColorEmoji.ttf", 109)

draw.text((30, 20), "Hello / Привет / Γειά", font=latin, fill="#0f172a")
draw.text((730, 90), "مرحبا بالعالم", font=latin, fill="#0f172a", anchor="ra")   # right-to-left, shaped
draw.text((30, 160), "你好，世界", font=cjk, fill="#0f172a")

# Colour emoji fonts are bitmaps at one size: load at 109, draw, then scale the layer.
e = Image.new("RGBA", (400, 140), (0, 0, 0, 0))
ImageDraw.Draw(e).text((0, 0), "🎨🐍📷", font=emoji, embedded_color=True)
e = e.resize((e.width * 60 // 109, e.height * 60 // 109), Image.LANCZOS)
img.paste(e, (330, 230), e)
img.save("scripts.png")
Tested on ✓ Linux (Linux-specific example)
Output
raqm layout engine available: True
Text in Latin, Cyrillic, Greek, Arabic and Chinese, plus three colour emoji, drawn with Pillow
Real output: DejaVu Sans for Latin, Cyrillic, Greek and Arabic, WenQuanYi Zen Hei for Chinese, Noto Color Emoji for emoji.

Colour emoji need embedded_color=True. Noto Color Emoji is a bitmap font that only exists at size 109, so any other size fails:

emoji.py
from PIL import ImageFont

ImageFont.truetype("/usr/share/fonts/truetype/noto/NotoColorEmoji.ttf", 48)
Tested on ✓ Linux (Linux-specific example)
Traceback
OSError: invalid pixel size

Load it at 109, draw the emoji on a transparent layer, and resize that layer, as in the script above.

Saving the result

Error'ImageDraw' object has no attribute 'textsize' ErrorOSError: cannot open resource (fonts) ToolWatermark generator

The outputs and images on this page are the real results of running every example with Pillow 12.3.0 on Python 3.13 (Linux) on 2026-10-15. Every example is also re-run automatically on macOS (Darwin 25.6.0, Pillow 12.3.0), Linux (Linux 6.17.0-1022-azure, Pillow 12.3.0), Windows (Windows 2025Server, Pillow 12.3.0); the line under each example shows the result per operating system. The DejaVu Sans font files sit next to each script so font examples behave the same everywhere. Sample photos: NASA/Sam Lott, public domain. Spotted something wrong? Tell us.