A picture was decoded at the size of its box on screen. A cropped picture
shows only part of the image, magnified, so it was decoded at a fraction of
the resolution it needs and stretched back up: Merge Shapes > Intersect,
which crops the picture to the new smaller box, left it visibly blurred.
Size the decode from the rectangle the whole image occupies instead.
A shape's picture fill marked not to rotate with the shape turned and
flipped with it (rotWithShape was neither kept nor written: always 1).
PowerPoint keeps such a picture upright, stretched (or tiled) over the
box the rotated shape covers on the slide and clipped to the shape. Now
read, written (PictureFill::upright) and drawn so.
Wild synthetic set: picture fills on rotated/flipped shapes 12.89 ->
3.87.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A gradient fill marked not to rotate with its shape turned and flipped
with it anyway. PowerPoint lays such a gradient out upright, over the box
the rotated or flipped shape covers on the slide (a 90 degree gradient on
a quarter-turned triangle still runs top to bottom). Linear and radial
gradients now do; rect and outline path gradients still follow the shape.
Wild synthetic set: gradients on rotated/flipped shapes 5.48 -> 1.34,
gradient angles slide 6.75 -> 1.95.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- flip x/y/xy (every other tile mirrored) was ignored: tiles now repeat
reflected across the flip axes.
- algn was ignored: the first tile now sits on the corner, edge middle or
centre of the box it names, before the tx/ty offset.
- The tile size came from the decoded (possibly downsampled) bitmap; it is
now the picture's own pixel size (96 dpi) times sx/sy, rounded to whole
output pixels as PowerPoint draws it.
- A scale of zero or less drew tiles at 100%; PowerPoint draws nothing.
- Tiles are one repeating image paint instead of one draw per tile (no
200-tile cap, constant cost).
Wild synthetic set: tiled background flipped/aligned 82.92 -> 1.80, tile
variants 51.15 -> 8.82.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Gradient tile rectangles were dropped on import. PowerPoint's "from
corner" radial presets depend on them: fillToRect puts the focus on a
corner and tileRect stretches the gradient's box a whole size past it, so
the circle is centred on that corner and reaches the far one. Without it
(after the two-point radial change) such fills washed out. A linear
gradient now spans its tile rectangle; a radial one ends on the circle
around it, with the focus still placed in the shape's own box (fitted to
PowerPoint's export of the corpus deck tdf114848: mean diff over the
filled box 0.77). Round-trips through pptx.
Wild synthetic set: tile-rect gradients 16.74 -> 2.48.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
An off-centre radial gradient ("from corner", or any fillToRect focus
other than the middle) spread from its focus out to the farthest corner,
so the near side stayed light. Fitted to PowerPoint's export: the first
stop is a point at the focus and the last is the circle through the box's
corners (centre of the box, half its diagonal), with the circles between
moving from one to the other (a two-point radial). A corner focus thus
reaches its last colour at its own corner too. Centred gradients are
unchanged.
Wild synthetic set: off-centre radial background 20.66 -> 1.22, radial
focus grid 6.34 -> 5.69.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Path gradients that follow a rectangle (path="rect") or the shape's own
outline (path="shape") were drawn as circles. They're now rasterised into
a small image of the bounds: rect takes the larger of the horizontal and
vertical shares of the distance from the focus to the edge (X-shaped
contours), shape the share of the way from the focus to the outline along
the same ray (an ellipse gives ellipses, a triangle triangles), as
measured on PowerPoint's export. The two path probe decks now match it to
a mean pixel difference of 1.7 and 0.6 (from 15.6 and 22).
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Pattern fills were approximated (dots, bars or a grid per family, on an
8-point cell), so pct30 looked like sparse dots and wave like a dot grid.
Every one of the 54 DrawingML presets now draws its 8x8 tile, each cell one
96-dpi pixel (0.75 pt), laid out from the slide's origin. The tiles were
read from PowerPoint's own export of each preset at two output pixels per
cell; the pattern probe deck now matches it to a mean pixel difference of
0.2 (from 62).
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
A gradient on text was read but drawn in its first colour. Each line is now
filled with the whole gradient, spanning that line's text from its top to
its bottom and its first to its last character, as PowerPoint does. The
gradient builder moves out of fill_path into paint::gradient so fills and
text share it (and PowerPoint's two-colour ramp). Gradient colours are
resolved when the text style is made.
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
PowerPoint blends a gradient of exactly two stops at 0% and 100% in linear
light, with the second colour's share following an S-shaped curve (black to
white is 105/186/238 at a quarter/half/three quarters, not 64/128/191).
Gradients with more stops, or a stop moved off an end, stay linear in sRGB.
Measured with PowerPoint's slide export on shapes and backgrounds, in both
directions. The two-colour case is expanded into sixteen pieces per
segment so vello's sRGB interpolation draws the same ramp.
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Theme background pictures (Office 2007-era themes such as Treck) carry an
a:duotone recolour in phClr; DeckCraft read it but drew the picture
unchanged, so those backgrounds came out grey. The painter now resolves the
duotone's colours against the slide's colours and placeholder colour, and
the decoded-image cache applies them (luminance picks a colour between dark
and light) and keys on them.
Layouts and slides can replace the master's colour map
(p:clrMapOvr/a:overrideClrMapping, e.g. dark section layouts mapping bg1 to
dk1). They were read as the master's map; now Layout and Slide carry an
optional colour map, resolve::Ctx uses slide > layout > master, and the
writer keeps them.
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
* feat(pictures): animated GIFs play, with pause (media.gifPlay)
GIF pictures used to show only their first frame. The renderer now decodes every frame (cached, capped at 256 MB and 1024 px a side) and draws the frame for each GIF's play time, passed in RenderOpts::gif_times; exports and thumbnails pass none and keep frame 1. The editor canvas and the slide show redraw only when a GIF reaches its next frame. A corner button on hover (or selection) pauses and resumes a GIF from the same frame, through the new media.gifPlay command.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(pictures): animated GIFs play, with pause (media.gifPlay)
GIF pictures used to show only their first frame. The renderer now decodes every frame (cached, capped at 256 MB and 1024 px a side) and draws the frame for each GIF's play time, passed in RenderOpts::gif_times; exports and thumbnails pass none and keep frame 1. The editor canvas and the slide show redraw only when a GIF reaches its next frame. A corner button on hover (or selection) pauses and resumes a GIF from the same frame, through the new media.gifPlay command.
Also adds my entry to contributors/people.toml.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Animated GIFs: cap frames (1000), canvas (4096²), per-GIF (64 MB) and cache (192 MB) memory
A few MB of hostile GIF could decode to millions of tiny frames or a 100 Mpx canvas per frame,
and eight cached GIFs could hold 2 GB.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
