[FFmpeg-devel] [PATCH] fix pad artifacting

[Ben Lu]

Sorry about that, it looks correct in gmail and sending to myself also
looks fine, not sure where the line width corruption is happening. Our org
doesn't allow programmatic email.

Is it possible to use this gist?
https://gist.github.com/ayroblu/1d7bbfdf82aa9a127abefa4c5d518cdb

Also trying again:

Signed-off-by: Ben Lu <benlu at x.com>
---
 libavfilter/vf_pad.c | 29 +++++++++++++++++++++--------
 1 file changed, 21 insertions(+), 8 deletions(-)

diff --git a/libavfilter/vf_pad.c b/libavfilter/vf_pad.c
index 49fb272b24..846dd0a9f4 100644
--- a/libavfilter/vf_pad.c
+++ b/libavfilter/vf_pad.c
@@ -237,15 +237,24 @@ static AVFrame *get_video_buffer(AVFilterLink
*inlink, int w, int h)
     if (s->inlink_w <= 0)
         return NULL;

-    frame = ff_get_video_buffer(inlink->dst->outputs[0],
-                                w + (s->w - s->in_w),
-                                h + (s->h - s->in_h) + (s->x > 0));
+    if (s->eval_mode == EVAL_MODE_FRAME) {
+        frame = ff_get_video_buffer(inlink->dst->outputs[0], s->w, s->h);
+    } else {
+        frame = ff_get_video_buffer(inlink->dst->outputs[0],
+                                    w + (s->w - s->in_w),
+                                    h + (s->h - s->in_h) + (s->x > 0));
+    }

     if (!frame)
         return NULL;

-    frame->width  = w;
-    frame->height = h;
+    if (s->eval_mode == EVAL_MODE_FRAME) {
+        frame->width  = s->w;
+        frame->height = s->h;
+    } else {
+        frame->width  = w;
+        frame->height = h;
+    }

     for (plane = 0; plane < 4 && frame->data[plane] &&
frame->linesize[plane]; plane++) {
         int hsub = s->draw.hsub[plane];
@@ -358,9 +367,13 @@ static int filter_frame(AVFilterLink *inlink, AVFrame
*in)

     if (needs_copy) {
         av_log(inlink->dst, AV_LOG_DEBUG, "Direct padding impossible
allocating new frame\n");
-        out = ff_get_video_buffer(outlink,
-                                  FFMAX(inlink->w, s->w),
-                                  FFMAX(inlink->h, s->h));
+        if (s->eval_mode == EVAL_MODE_FRAME) {
+            out = ff_get_video_buffer(outlink, s->w, s->h);
+        } else {
+            out = ff_get_video_buffer(outlink,
+                                      FFMAX(inlink->w, s->w),
+                                      FFMAX(inlink->h, s->h));
+        }
         if (!out) {
             av_frame_free(&in);
             return AVERROR(ENOMEM);
-- 
2.35.1.70.gdb80f58b59-twtrsrc

On Sat, Feb 1, 2025 at 12:52 AM Michael Niedermayer <michael at niedermayer.cc>
wrote:

> Hi
>
> On Sat, Feb 01, 2025 at 12:18:08AM +0000, Ben Lu via ffmpeg-devel wrote:
> > When using pad with eval=frame with variable frame sizes, we get
> > significant artifacting. This is due to incorrect frame sizes, resulting
> in
> > invalid frames. Made changes to use the output width and height for frame
> > sizes when using eval=frame, with if statement guards to make sure the
> > normal usage is unaffected
> >
> > Signed-off-by: Ben Lu <benlu at x.com>
> > ---
> >  libavfilter/vf_pad.c | 29 +++++++++++++++++++++--------
> >  1 file changed, 21 insertions(+), 8 deletions(-)
> >
> > diff --git a/libavfilter/vf_pad.c b/libavfilter/vf_pad.c
> > index 49fb272b24..846dd0a9f4 100644
> > --- a/libavfilter/vf_pad.c
> > +++ b/libavfilter/vf_pad.c
> > @@ -237,15 +237,24 @@ static AVFrame *get_video_buffer(AVFilterLink
> > *inlink, int w, int h)
>
> error: corrupt patch at line 10
>
> check your editor line/word wrap break settings or attach the patch or use
> git send-email
>
> thx
>
> [...]
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> "I am not trying to be anyone's saviour, I'm trying to think about the
>  future and not be sad" - Elon Musk
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>