diff --git a/libavcodec/ffv1.c b/libavcodec/ffv1.c
index cbb1fe8fbc..1b038b11c9 100644
--- a/libavcodec/ffv1.c
+++ b/libavcodec/ffv1.c
@@ -49,8 +49,8 @@ av_cold int ffv1_common_init(AVCodecContext *avctx)
     s->avctx = avctx;
     s->flags = avctx->flags;
 
-    avcodec_get_frame_defaults(&s->picture);
-
+    s->picture.f = avcodec_alloc_frame();
+    s->last_picture.f = av_frame_alloc();
     ff_dsputil_init(&s->dsp, avctx);
 
     s->width  = avctx->width;
@@ -192,7 +192,13 @@ av_cold int ffv1_close(AVCodecContext *avctx)
     FFV1Context *s = avctx->priv_data;
     int i, j;
 
-    av_frame_unref(&s->last_picture);
+    if (s->picture.f)
+        ff_thread_release_buffer(avctx, &s->picture);
+    av_frame_free(&s->picture.f);
+
+    if (s->last_picture.f)
+        ff_thread_release_buffer(avctx, &s->last_picture);
+    av_frame_free(&s->last_picture.f);
 
     for (j = 0; j < s->slice_count; j++) {
         FFV1Context *fs = s->slice_context[j];
diff --git a/libavcodec/ffv1.h b/libavcodec/ffv1.h
index 981004798e..781ddf2bea 100644
--- a/libavcodec/ffv1.h
+++ b/libavcodec/ffv1.h
@@ -41,6 +41,7 @@
 #include "mathops.h"
 #include "put_bits.h"
 #include "rangecoder.h"
+#include "thread.h"
 
 #ifdef __INTEL_COMPILER
 #undef av_flatten
@@ -89,7 +90,7 @@ typedef struct FFV1Context {
     int transparency;
     int flags;
     int picture_number;
-    AVFrame picture, last_picture;
+    ThreadFrame picture, last_picture;
 
     AVFrame *cur;
     int plane_count;
diff --git a/libavcodec/ffv1dec.c b/libavcodec/ffv1dec.c
index d73c2e04e6..d623233ea8 100644
--- a/libavcodec/ffv1dec.c
+++ b/libavcodec/ffv1dec.c
@@ -38,7 +38,6 @@
 #include "golomb.h"
 #include "mathops.h"
 #include "ffv1.h"
-#include "thread.h"
 
 static inline av_flatten int get_symbol_inline(RangeCoder *c, uint8_t *state,
                                                int is_signed)
@@ -734,14 +733,16 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPac
     int buf_size        = avpkt->size;
     FFV1Context *f      = avctx->priv_data;
     RangeCoder *const c = &f->slice_context[0]->c;
-    ThreadFrame frame = { .f = data };
     int i, ret;
     uint8_t keystate = 128;
     const uint8_t *buf_p;
-    AVFrame *const p    = data;
+    AVFrame *p;
 
-    f->cur = p;
-    f->avctx = avctx;
+    if (f->last_picture.f)
+        ff_thread_release_buffer(avctx, &f->last_picture);
+    FFSWAP(ThreadFrame, f->picture, f->last_picture);
+
+    f->cur = p = f->picture.f;
 
     ff_init_range_decoder(c, buf, buf_size);
     ff_build_rac_states(c, 0.05 * (1LL << 32), 256 - 8);
@@ -762,7 +763,7 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPac
         p->key_frame = 0;
     }
 
-    if ((ret = ff_thread_get_buffer(avctx, &frame, AV_GET_BUFFER_FLAG_REF)) < 0)
+    if ((ret = ff_thread_get_buffer(avctx, &f->picture, AV_GET_BUFFER_FLAG_REF)) < 0)
         return ret;
 
     if (avctx->debug & FF_DEBUG_PICT_INFO)
@@ -818,7 +819,7 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPac
     for (i = f->slice_count - 1; i >= 0; i--) {
         FFV1Context *fs = f->slice_context[i];
         int j;
-        if (fs->slice_damaged && f->last_picture.data[0]) {
+        if (fs->slice_damaged && f->last_picture.f->data[0]) {
             const uint8_t *src[4];
             uint8_t *dst[4];
             for (j = 0; j < 4; j++) {
@@ -826,11 +827,11 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPac
                 int sv = (j==1 || j==2) ? f->chroma_v_shift : 0;
                 dst[j] = p->data[j] + p->linesize[j]*
                          (fs->slice_y>>sv) + (fs->slice_x>>sh);
-                src[j] = f->last_picture.data[j] + f->last_picture.linesize[j]*
+                src[j] = f->last_picture.f->data[j] + f->last_picture.f->linesize[j]*
                          (fs->slice_y>>sv) + (fs->slice_x>>sh);
             }
             av_image_copy(dst, p->linesize, (const uint8_t **)src,
-                          f->last_picture.linesize,
+                          f->last_picture.f->linesize,
                           avctx->pix_fmt,
                           fs->slice_width,
                           fs->slice_height);
@@ -839,10 +840,11 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, AVPac
 
     f->picture_number++;
 
-    av_frame_unref(&f->last_picture);
-    if ((ret = av_frame_ref(&f->last_picture, p)) < 0)
-        return ret;
+    if (f->last_picture.f)
+        ff_thread_release_buffer(avctx, &f->last_picture);
     f->cur = NULL;
+    if ((ret = av_frame_ref(data, f->picture.f)) < 0)
+        return ret;
 
     *got_frame = 1;
 
diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c
index 828ed9b6e0..79e483d6dd 100644
--- a/libavcodec/ffv1enc.c
+++ b/libavcodec/ffv1enc.c
@@ -794,7 +794,7 @@ static av_cold int encode_init(AVCodecContext *avctx)
     if ((ret = ffv1_allocate_initial_states(s)) < 0)
         return ret;
 
-    avctx->coded_frame = &s->picture;
+    avctx->coded_frame = s->picture.f;
     if (!s->transparency)
         s->plane_count = 2;
     avcodec_get_chroma_sub_sample(avctx->pix_fmt, &s->chroma_h_shift, &s->chroma_v_shift);
@@ -939,12 +939,12 @@ static void encode_slice_header(FFV1Context *f, FFV1Context *fs)
         put_symbol(c, state, f->plane[j].quant_table_index, 0);
         av_assert0(f->plane[j].quant_table_index == f->avctx->context_model);
     }
-    if (!f->picture.interlaced_frame)
+    if (!f->picture.f->interlaced_frame)
         put_symbol(c, state, 3, 0);
     else
-        put_symbol(c, state, 1 + !f->picture.top_field_first, 0);
-    put_symbol(c, state, f->picture.sample_aspect_ratio.num, 0);
-    put_symbol(c, state, f->picture.sample_aspect_ratio.den, 0);
+        put_symbol(c, state, 1 + !f->picture.f->top_field_first, 0);
+    put_symbol(c, state, f->picture.f->sample_aspect_ratio.num, 0);
+    put_symbol(c, state, f->picture.f->sample_aspect_ratio.den, 0);
 }
 
 static int encode_slice(AVCodecContext *c, void *arg)
@@ -955,7 +955,7 @@ static int encode_slice(AVCodecContext *c, void *arg)
     int height       = fs->slice_height;
     int x            = fs->slice_x;
     int y            = fs->slice_y;
-    AVFrame *const p = &f->picture;
+    AVFrame *const p = f->picture.f;
     const int ps     = av_pix_fmt_desc_get(c->pix_fmt)->comp[0].step_minus1 + 1;
 
     if (p->key_frame)
@@ -1002,7 +1002,7 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
 {
     FFV1Context *f      = avctx->priv_data;
     RangeCoder *const c = &f->slice_context[0]->c;
-    AVFrame *const p    = &f->picture;
+    AVFrame *const p    = f->picture.f;
     int used_count      = 0;
     uint8_t keystate    = 128;
     uint8_t *buf_p;
@@ -1015,7 +1015,7 @@ static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
     ff_init_range_encoder(c, pkt->data, pkt->size);
     ff_build_rac_states(c, 0.05 * (1LL << 32), 256 - 8);
 
-    *p           = *pict;
+    av_frame_ref(p, pict);
     p->pict_type = AV_PICTURE_TYPE_I;
 
     if (avctx->gop_size == 0 || f->picture_number % avctx->gop_size == 0) {