Add camu

- Subprojects temporarily omitted

Signed-off-by: Andrew Opalach <andrew@akon.city>

16 files changed, 1130 insertions, 0 deletions

diff --git a/src/codec/libav/avio.c b/src/codec/libav/avio.c
new file mode 100644
index 0000000..97f66da
--- /dev/null
+++ b/src/codec/libav/avio.c
@@ -0,0 +1,13 @@
+#include "avio.h"
+
+s32 camu_avio_read(void *data, u8 *buf, s32 buf_size)
+{
+    struct cch_handle *handle = (struct cch_handle *)data;
+    return cch_handle_read(handle, buf, buf_size);
+}
+
+s64 camu_avio_seek(void *data, s64 offset, s32 whence)
+{
+    struct cch_handle *handle = (struct cch_handle *)data;
+    return (s64)cch_handle_seek(handle, offset, whence);
+}
diff --git a/src/codec/libav/avio.h b/src/codec/libav/avio.h
new file mode 100644
index 0000000..09df4ec
--- /dev/null
+++ b/src/codec/libav/avio.h
@@ -0,0 +1,10 @@
+#pragma once
+
+#include <libavutil/error.h>
+#include <libavformat/avio.h>
+
+#include "../../codec/codec.h"
+#include "../../cache/handle.h"
+
+s32 camu_avio_read(void *data, u8 *buf, s32 buf_size);
+s64 camu_avio_seek(void *data, s64 offset, s32 whence);
diff --git a/src/codec/libav/common.c b/src/codec/libav/common.c
new file mode 100644
index 0000000..8ce6667
--- /dev/null
+++ b/src/codec/libav/common.c
@@ -0,0 +1,6 @@
+#include "common.h"
+
+void camu_lav_set_log_callback(void (*callback)(void *, int, const char *, va_list))
+{
+    av_log_set_callback(callback);
+}
diff --git a/src/codec/libav/common.h b/src/codec/libav/common.h
new file mode 100644
index 0000000..854e4c7
--- /dev/null
+++ b/src/codec/libav/common.h
@@ -0,0 +1,6 @@
+#pragma once
+
+#include <al/types.h>
+#include <libavutil/log.h>
+
+void camu_lav_set_log_callback(void (*callback)(void *, int, const char *, va_list));
diff --git a/src/codec/libav/decoder.c b/src/codec/libav/decoder.c
new file mode 100644
index 0000000..ebdcb03
--- /dev/null
+++ b/src/codec/libav/decoder.c
@@ -0,0 +1,165 @@
+#include <libavutil/cpu.h>
+#include <al/lib.h>
+#include <al/log.h>
+
+//#include "../../render/renderer.h"
+
+#include "decoder.h"
+
+static void close_internal(struct camu_lav_decoder *av)
+{
+    if (av->codec_context) {
+        avcodec_close(av->codec_context);
+        avcodec_free_context(&av->codec_context);
+    }
+}
+
+static bool lav_decoder_init(struct camu_decoder *dec, struct camu_renderer *renderer, struct camu_stream *stream)
+{
+    struct camu_lav_decoder *av = (struct camu_lav_decoder *)dec;
+
+    AVCodecParameters *codecpar = stream->av.stream->codecpar;
+    const AVCodec *codec = avcodec_find_decoder(codecpar->codec_id);
+
+    if (!codec) {
+        al_log_error("lav_decoder", "Failed to find decoder.");
+        goto err;
+    }
+
+    av->codec_context = avcodec_alloc_context3(codec);
+
+    if (!av->codec_context) {
+        al_log_error("lav_decoder", "Failed to alloc codec context.");
+        goto err;
+    }
+
+    if (avcodec_parameters_to_context(av->codec_context, codecpar) < 0) {
+        al_log_error("lav_decoder", "Failed to copy codec parameters.");
+        goto err;
+    }
+
+    s32 cpus = av_cpu_count();
+    cpus = (cpus > 16) ? 4 : cpus / 4;
+    if (cpus == 0) cpus = 1;
+    av->codec_context->thread_count = cpus;
+    al_log_debug("lav_decoder", "Using %i threads for decoder.", cpus);
+
+    (void)renderer;
+    /*
+    if (renderer && renderer->get_buffer2 && codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {
+        av->codec_context->get_buffer2 = renderer->get_buffer2;
+        av->codec_context->opaque = renderer->opaque;
+    }
+    */
+
+    if (avcodec_open2(av->codec_context, codec, NULL) < 0) {
+        al_log_error("lav_decoder", "Failed to open codec (%s).", codec->name);
+        goto err;
+    }
+
+    al_log_info("lav_decoder", "Codec: %s (%s) %ldkbps.", codec->name, codec->long_name, codecpar->bit_rate / 1000);
+
+    av->time_base = stream->av.stream->time_base;
+    av->last_pts = 0;
+    av->last_duration = 0;
+    //av->seek_pos = -1;
+
+    return true;
+err:
+    close_internal(av);
+
+    return false;
+}
+
+static void lav_decoder_set_callback(struct camu_decoder *dec, void (*callback)(void *, struct camu_frame *), void *userdata)
+{
+    struct camu_lav_decoder *av = (struct camu_lav_decoder *)dec;
+    av->callback = callback;
+    av->userdata = userdata;
+}
+
+static s32 send_packet(struct camu_lav_decoder *av, AVPacket *pkt)
+{
+    /*
+    if (pkt && av->seek_pos >= 0 && pkt->pts + pkt->duration < av->seek_pos) {
+        av->codec_context->skip_frame = AVDISCARD_NONKEY;
+    } else {
+        av->codec_context->skip_frame = AVDISCARD_NONE;
+    }
+    */
+
+    s32 ret = avcodec_send_packet(av->codec_context, pkt);
+    if (ret < 0 && ret != AVERROR(EAGAIN) && ret != AVERROR_EOF) {
+        al_log_error("lav_decoder", "Error sending packet to decoder: (%s).", av_err2str(ret));
+    }
+
+    return ret;
+}
+
+static s32 lav_decoder_push(struct camu_decoder *dec, struct camu_packet *packet)
+{
+    struct camu_lav_decoder *av = (struct camu_lav_decoder *)dec;
+    return send_packet(av, (packet) ? packet->av.pkt : NULL);
+}
+
+static void lav_decoder_flush(struct camu_decoder *dec)
+{
+    struct camu_lav_decoder *av = (struct camu_lav_decoder *)dec;
+    avcodec_flush_buffers(av->codec_context);
+}
+
+static s32 receive_frames(struct camu_lav_decoder *av)
+{
+    s32 ret;
+    do {
+        struct camu_frame *frame = al_alloc_object(struct camu_frame);
+        frame->type = CAMU_FFMPEG_COMPAT;
+        frame->av.frame = av_frame_alloc();
+        ret = avcodec_receive_frame(av->codec_context, frame->av.frame);
+        if (ret < 0) {
+            av_frame_free(&frame->av.frame);
+            al_free(frame);
+            if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) break;
+            al_log_error("lav_decoder", "Error receiving packet from decoder: (%s).", av_err2str(ret));
+            continue;
+        }
+        // Track pts and duration of the previous frame so we can handle multiple frames
+        // in a single packet.
+        //if (av->codec_context->codec_type == AVMEDIA_TYPE_AUDIO) {
+        //    if (frame->av.frame->best_effort_timestamp < 0) {
+        //        frame->av.frame->best_effort_timestamp = av->last_pts + av->last_duration;
+        //    }
+        //    av->last_pts = frame->av.frame->best_effort_timestamp;
+        //    av->last_duration = av_get_audio_frame_duration(av->codec_context, frame->av.frame->linesize[0])
+        //      / (av->codec_context->sample_rate * av->time_base.num / (f64)av->time_base.den);
+        //}
+        av->callback(av->userdata, frame);
+    } while (1);
+    return ret;
+}
+
+static s32 lav_decoder_process(struct camu_decoder *dec)
+{
+    struct camu_lav_decoder *av = (struct camu_lav_decoder *)dec;
+    return receive_frames(av);
+}
+
+static void lav_decoder_free(struct camu_decoder **dec)
+{
+    struct camu_lav_decoder *av = (struct camu_lav_decoder *)*dec;
+    close_internal(av);
+    al_free(av);
+    *dec = NULL;
+}
+
+struct camu_decoder *camu_lav_decoder_create(void)
+{
+    struct camu_lav_decoder *av = al_alloc_object(struct camu_lav_decoder);
+    av->dec.init = lav_decoder_init;
+    av->dec.set_callback = lav_decoder_set_callback;
+    av->dec.push = lav_decoder_push;
+    av->dec.process = lav_decoder_process;
+    av->dec.flush = lav_decoder_flush;
+    av->dec.free = lav_decoder_free;
+    return (struct camu_decoder *)av;
+}
diff --git a/src/codec/libav/decoder.h b/src/codec/libav/decoder.h
new file mode 100644
index 0000000..44105ce
--- /dev/null
+++ b/src/codec/libav/decoder.h
@@ -0,0 +1,20 @@
+#pragma once
+
+#include <al/types.h>
+
+#include <libavcodec/avcodec.h>
+
+#include "../codec.h"
+
+struct camu_lav_decoder {
+    struct camu_decoder dec;
+    AVCodecContext *codec_context;
+    AVRational time_base;
+    s64 last_pts;
+    s64 last_duration;
+    //s64 seek_pos;
+    void (*callback)(void *, struct camu_frame *);
+    void *userdata;
+};
+
+struct camu_decoder *camu_lav_decoder_create(void);
diff --git a/src/codec/libav/demuxer.c b/src/codec/libav/demuxer.c
new file mode 100644
index 0000000..f7ff35c
--- /dev/null
+++ b/src/codec/libav/demuxer.c
@@ -0,0 +1,184 @@
+#include <al/log.h>
+
+#include "../common.h"
+
+#include "demuxer.h"
+
+#define BUF_SIZE 16384
+
+static void close_internal(struct camu_lav_demuxer *av)
+{
+    if (av->io_context) {
+        av_free(av->io_context->buffer);
+        avio_context_free(&av->io_context);
+    }
+    if (av->format_context) {
+        avformat_flush(av->format_context);
+        avformat_close_input(&av->format_context);
+        avformat_free_context(av->format_context);
+    }
+}
+
+static bool lav_demuxer_init(struct camu_demuxer *demux, struct cch_handle *handle)
+{
+    struct camu_lav_demuxer *av = (struct camu_lav_demuxer *)demux;
+
+    av->format_context = avformat_alloc_context();
+    if (!av->format_context) {
+        al_log_error("lav_demuxer", "Failed to create format context.");
+        goto err;
+    }
+
+    u8 *buf = (u8 *)av_malloc(BUF_SIZE);
+    av->io_context = avio_alloc_context(buf, BUF_SIZE, 0, handle, camu_avio_read, NULL, camu_avio_seek);
+    if (!av->io_context) {
+        al_log_error("lav_demuxer", "Failed to create custom io context.");
+        goto err;
+    }
+
+    if (!cch_handle_can_seek(handle)) {
+        // This should always be unset if we know we cannot seek.
+        // But, should also never be unset if we *do* want to seek because it
+        // disables long seeks (Not tested in a long time).
+        av->io_context->seekable &= ~AVIO_SEEKABLE_NORMAL;
+    } else {
+        av->io_context->seekable |= AVIO_SEEKABLE_NORMAL;
+    }
+
+    av->format_context->pb = av->io_context;
+    av->format_context->flags |= AVFMT_FLAG_CUSTOM_IO;
+
+    if (avformat_open_input(&av->format_context, "", 0, 0) < 0) {
+        al_log_error("lav_demuxer", "Failed to open input.");
+        goto err;
+    }
+
+    if (avformat_find_stream_info(av->format_context, 0) < 0) {
+        al_log_error("lav_demux", "Failed to find stream info.");
+        goto err;
+    }
+
+    if (!av->format_context->nb_streams) {
+        al_log_error("lav_demux", "No streams found.");
+        goto err;
+    }
+
+    av->duration = 0;
+
+    bool audio_selected = false;
+    bool video_selected = false;
+
+    for (u32 i = 0; i < av->format_context->nb_streams; i++) {
+        AVStream *stream = av->format_context->streams[i];
+        enum AVMediaType type = stream->codecpar->codec_type;
+        if (!(type == AVMEDIA_TYPE_AUDIO || type == AVMEDIA_TYPE_VIDEO)) {
+            continue;
+        }
+        if (type == AVMEDIA_TYPE_AUDIO) {
+            if (audio_selected) continue;
+            else audio_selected = true;
+        }
+        if (type == AVMEDIA_TYPE_VIDEO) {
+            if (video_selected) continue;
+            else video_selected = true;
+        }
+        // Try to detect attached images.
+        if (type == AVMEDIA_TYPE_VIDEO && stream->duration >= 0 && stream->nb_frames <= 1 &&
+            (stream->avg_frame_rate.den == 0 && stream->r_frame_rate.den > 0)) {
+            stream->duration = 0;
+            al_log_info("lav_demux", "Assuming stream #%i is an image.", i);
+        } else if (stream->duration < 0) {
+            if (av->format_context->duration < 0) {
+                stream->duration = 0;
+            } else {
+                // This could be completely wrong. In that case, it's still probably
+                // better than 0 (assuming the stream is not an image).
+                stream->duration = av_rescale_q(av->format_context->duration,
+                    AV_TIME_BASE_Q, stream->time_base);
+            }
+            al_log_info("lav_demux", "Setting stream #%i to a duration of %.3fs.",
+                i, stream->duration * av_q2d(stream->time_base));
+        }
+        if (stream->start_time < 0) {
+            stream->start_time = 0;
+        }
+        s64 duration = av_rescale_q(stream->duration, stream->time_base, AV_TIME_BASE_Q);
+        if (duration > av->duration) {
+            av->duration = duration;
+        }
+        al_array_push(av->demux.streams, ((struct camu_stream){
+            .type = CAMU_FFMPEG_COMPAT,
+            .av.stream = stream
+        }));
+    }
+
+    av_format_inject_global_side_data(av->format_context);
+
+    //av_dump_format(av->format_context, 0, "", 0);
+
+    av->eof = false;
+
+    return true;
+err:
+    close_internal(av);
+
+    return false;
+}
+
+static s32 lav_demuxer_get_packet(struct camu_demuxer *demux, struct camu_packet *packet)
+{
+    struct camu_lav_demuxer *av = (struct camu_lav_demuxer *)demux;
+    packet->type = CAMU_FFMPEG_COMPAT;
+    s32 ret = 0;
+    do {
+        ret = av_read_frame(av->format_context, packet->av.pkt);
+        if (ret < 0 && ret != AVERROR_EOF) {
+            al_log_error("lav_demuxer", "Failed to read frame (%s).", av_err2str(ret));
+        }
+        break;
+    } while (1);
+    if (ret == AVERROR_EOF) av->eof = true;
+    return ret;
+}
+
+static s64 lav_demuxer_get_duration(struct camu_demuxer *demux)
+{
+    struct camu_lav_demuxer *av = (struct camu_lav_demuxer *)demux;
+    return av->duration;
+}
+
+static bool lav_demuxer_seek(struct camu_demuxer *demux, s64 pos)
+{
+    struct camu_lav_demuxer *av = (struct camu_lav_demuxer *)demux;
+    s64 ts = av_rescale_q(pos, (AVRational){ 1, 1000000 }, AV_TIME_BASE_Q);
+    if (av->eof) {
+        avformat_flush(av->format_context);
+        av->eof = false;
+    }
+    if (avformat_seek_file(av->format_context, -1, INT64_MIN, ts, ts, AVSEEK_FLAG_BACKWARD) >= 0) {
+        return true;
+    }
+    return false;
+}
+
+static void lav_demuxer_free(struct camu_demuxer **demux)
+{
+    struct camu_lav_demuxer *av = (struct camu_lav_demuxer *)*demux;
+    close_internal(av);
+    al_array_free(av->demux.streams);
+    al_free(av);
+    *demux = NULL;
+}
+
+struct camu_demuxer *camu_lav_demuxer_create(void)
+{
+    struct camu_lav_demuxer *av = al_alloc_object(struct camu_lav_demuxer);
+    av->demux.type = CAMU_FFMPEG_COMPAT;
+    al_array_init(av->demux.streams);
+    av->demux.init = lav_demuxer_init;
+    av->demux.get_packet = lav_demuxer_get_packet;
+    av->demux.get_duration = lav_demuxer_get_duration;
+    av->demux.seek = lav_demuxer_seek;
+    av->demux.free = lav_demuxer_free;
+    return (struct camu_demuxer *)av;
+}
diff --git a/src/codec/libav/demuxer.h b/src/codec/libav/demuxer.h
new file mode 100644
index 0000000..a9d8c5c
--- /dev/null
+++ b/src/codec/libav/demuxer.h
@@ -0,0 +1,21 @@
+#pragma once
+
+#include <al/types.h>
+#include <al/lib.h>
+#include <al/array.h>
+
+#include <libavformat/avformat.h>
+
+#include "../codec.h"
+
+#include "avio.h"
+
+struct camu_lav_demuxer {
+    struct camu_demuxer demux;
+    AVIOContext *io_context;
+    AVFormatContext *format_context;
+    s64 duration;
+    bool eof;
+};
+
+struct camu_demuxer *camu_lav_demuxer_create(void);
diff --git a/src/codec/libav/encoder.c b/src/codec/libav/encoder.c
new file mode 100644
index 0000000..4f3fa80
--- /dev/null
+++ b/src/codec/libav/encoder.c
@@ -0,0 +1,186 @@
+#include <libavcodec/avcodec.h>
+#include <libavcodec/codec.h>
+#include <libavformat/avformat.h>
+#include <libavutil/audio_fifo.h>
+#include <libavutil/channel_layout.h>
+#include <libavutil/error.h>
+#include <libavutil/frame.h>
+#include <libavutil/samplefmt.h>
+#include <al/lib.h>
+#include <al/log.h>
+
+#include "encoder.h"
+
+#define BITRATE 256000
+#define FRAME_SIZE 20
+#define CHANNELS 2
+
+bool camu_lav_encoder_init(struct camu_lav_encoder *enc, char *encoder_name, char *ext)
+{
+    al_memset(enc, 0, sizeof(struct camu_lav_encoder));
+
+    const AVCodec *codec = avcodec_find_encoder_by_name(encoder_name);
+
+    if (!codec) {
+        al_log_error("lav_encoder", "Failed to find codec with name: %s.", encoder_name);
+        goto err;
+    }
+
+    enc->codec_context = avcodec_alloc_context3(codec);
+
+    if (!enc->codec_context) {
+        al_log_error("lav_encoder", "Failed to alloc codec context.");
+        goto err;
+    }
+
+    enc->codec_context->sample_rate = 48000;
+    av_channel_layout_default(&enc->codec_context->ch_layout, CHANNELS);
+    enc->codec_context->sample_fmt = AV_SAMPLE_FMT_FLT;
+    enc->codec_context->bit_rate = BITRATE;
+    enc->codec_context->frame_size = FRAME_SIZE;
+    enc->codec_context->time_base = (AVRational){ 1, enc->codec_context->sample_rate };
+    enc->codec_context->strict_std_compliance = FF_COMPLIANCE_EXPERIMENTAL;
+
+    enc->output_format = av_guess_format(ext, NULL, NULL);
+    if (enc->output_format && enc->output_format->flags & AVFMT_GLOBALHEADER) {
+        enc->codec_context->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
+    }
+
+    AVDictionary *opts = NULL;
+    av_dict_set(&opts, "b", "256k", 0);
+    av_dict_set(&opts, "vbr", "off", 0);
+    av_dict_set(&opts, "compression_level", "10", 0);
+    av_dict_set(&opts, "frame_duration", "20", 0);
+    av_dict_set(&opts, "application", "audio", 0);
+
+    if (avcodec_open2(enc->codec_context, codec, &opts) < 0) {
+        al_log_error("lav_encoder", "Failed to open codec.");
+        av_dict_free(&opts);
+        goto err;
+    }
+
+    av_dict_free(&opts);
+
+    AVCodecContext *out_codec = enc->codec_context;
+
+    av_channel_layout_copy(&enc->fmt.req_channel_layout, &out_codec->ch_layout);
+    enc->fmt.req_sample_rate = out_codec->sample_rate;
+    enc->fmt.req_format = out_codec->sample_fmt;
+
+    enc->audio_fifo = av_audio_fifo_alloc(enc->fmt.req_format,
+        enc->fmt.req_channel_layout.nb_channels, 1);
+
+    if (!enc->audio_fifo) {
+        al_log_error("lav_encoder", "Failed to alloc fifo.");
+        goto err;
+    }
+
+    enc->frame = av_frame_alloc();
+    enc->frame->nb_samples = 0;
+    enc->frame->sample_rate = enc->fmt.req_sample_rate;
+    av_channel_layout_copy(&enc->frame->ch_layout, &enc->fmt.req_channel_layout);
+    enc->frame->format = enc->fmt.req_format;
+    enc->frame->pts = 0;
+
+    return true;
+err:
+    camu_lav_encoder_close(enc);
+    return false;
+}
+
+void camu_lav_encoder_push(struct camu_lav_encoder *enc, u8 **data, s32 sample_count)
+{
+    s32 alloc_size = av_audio_fifo_size(enc->audio_fifo) + sample_count;
+    if (av_audio_fifo_space(enc->audio_fifo) < alloc_size) {
+        if (av_audio_fifo_realloc(enc->audio_fifo, alloc_size) < 0) {
+            al_assert(false);
+        }
+    }
+    av_audio_fifo_write(enc->audio_fifo, (void **)data, sample_count);
+}
+
+static void free_frame_data(struct camu_lav_encoder *enc)
+{
+    if (enc->frame && enc->frame->nb_samples > 0) av_freep(&enc->frame->data[0]);
+}
+
+static void alloc_frame_data(struct camu_lav_encoder *enc, s32 frame_size)
+{
+    if (enc->frame->nb_samples >= frame_size) return;
+    free_frame_data(enc);
+    s32 nb_channels = enc->frame->ch_layout.nb_channels;
+    av_samples_alloc(enc->frame->data, NULL, nb_channels, frame_size, enc->frame->format, 0);
+    enc->frame->nb_samples = frame_size;
+}
+
+static bool receive_packets(struct camu_lav_encoder *enc)
+{
+    s32 ret;
+    AVPacket pkt = { 0 };
+
+    do {
+        ret = avcodec_receive_packet(enc->codec_context, &pkt);
+        if (ret < 0) break;
+        enc->callback(enc->userdata, &pkt);
+    } while (1);
+
+    if (ret != AVERROR(EAGAIN) && ret != AVERROR_EOF) {
+        al_log_error("lav_encoder", "Error receiving packet from encoder: (%s).", av_err2str(ret));
+        return false;
+    }
+
+    return true;
+}
+
+static s32 send_frame(struct camu_lav_encoder *enc, AVFrame *in_frame)
+{
+    s32 ret = avcodec_send_frame(enc->codec_context, in_frame);
+    if (ret < 0 && ret != AVERROR_EOF) {
+        al_log_error("lav_encoder", "Error sending frame to encoder: (%s).", av_err2str(ret));
+    }
+    return ret;
+}
+
+bool camu_lav_encoder_process(struct camu_lav_encoder *enc, bool flush)
+{
+    s32 samples_available = av_audio_fifo_size(enc->audio_fifo);
+
+    if (samples_available == 0 && flush) {
+        // Flush encoder.
+        send_frame(enc, NULL);
+        receive_packets(enc);
+        // This should guarantee that there are not packets left.
+        return false;
+    }
+
+    s32 frame_size = enc->codec_context->frame_size;
+
+    if (samples_available == 0 || (samples_available < frame_size && !flush)) {
+        return false;
+    }
+
+    if (samples_available < frame_size) frame_size = samples_available;
+
+    alloc_frame_data(enc, frame_size);
+    av_audio_fifo_read(enc->audio_fifo, (void **)enc->frame->data, frame_size);
+
+    s32 ret;
+    do {
+        ret = send_frame(enc, enc->frame);
+        // TODO: Check ret < 0 ?
+        if (!receive_packets(enc)) return false;
+    } while (ret == AVERROR(EAGAIN));
+
+    enc->frame->pts += frame_size;
+
+    // ret < 0 means encoder erroed while sending frame.
+    return ret == 0;
+}
+
+void camu_lav_encoder_close(struct camu_lav_encoder *enc)
+{
+    if (enc->codec_context) avcodec_free_context(&enc->codec_context);
+    if (enc->audio_fifo) av_audio_fifo_free(enc->audio_fifo);
+    free_frame_data(enc);
+    if (enc->frame) av_frame_free(&enc->frame);
+}
diff --git a/src/codec/libav/encoder.h b/src/codec/libav/encoder.h
new file mode 100644
index 0000000..b5a707a
--- /dev/null
+++ b/src/codec/libav/encoder.h
@@ -0,0 +1,26 @@
+#pragma once
+
+#include <libavcodec/avcodec.h>
+#include <libavformat/avformat.h>
+#include <libavutil/audio_fifo.h>
+#include <libavutil/rational.h>
+
+#include "resampler.h"
+
+struct camu_lav_encoder {
+    AVCodecContext *codec_context;
+    const AVOutputFormat *output_format;
+
+    struct camu_lav_resample_fmt fmt;
+
+    AVFrame *frame;
+    AVAudioFifo *audio_fifo;
+
+    void (*callback)(void *, AVPacket *);
+    void *userdata;
+};
+
+bool camu_lav_encoder_init(struct camu_lav_encoder *enc, char *encoder_name, char *ext);
+void camu_lav_encoder_push(struct camu_lav_encoder *enc, u8 **data, s32 sample_count);
+bool camu_lav_encoder_process(struct camu_lav_encoder *enc, bool flush);
+void camu_lav_encoder_close(struct camu_lav_encoder *enc);
diff --git a/src/codec/libav/packet_ext.c b/src/codec/libav/packet_ext.c
new file mode 100644
index 0000000..8fe4c39
--- /dev/null
+++ b/src/codec/libav/packet_ext.c
@@ -0,0 +1,168 @@
+#include "packet_ext.h"
+
+void aki_packet_write_av_codec_parameters(struct aki_packet *packet, AVCodecParameters *codecpar)
+{
+    AKI_PACKET_WRITE_TYPE(packet, enum AVMediaType, codecpar->codec_type);
+    AKI_PACKET_WRITE_TYPE(packet, enum AVCodecID, codecpar->codec_id);
+    AKI_PACKET_WRITE_TYPE(packet, u32, codecpar->codec_tag);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->extradata_size);
+    AKI_PACKET_WRITE_DATA(packet, codecpar->extradata, codecpar->extradata_size);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->format);
+    AKI_PACKET_WRITE_TYPE(packet, s64, codecpar->bit_rate);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->bits_per_coded_sample);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->bits_per_raw_sample);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->profile);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->level);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->width);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->height);
+    AKI_PACKET_WRITE_TYPE(packet, AVRational, codecpar->sample_aspect_ratio);
+    AKI_PACKET_WRITE_TYPE(packet, enum AVFieldOrder, codecpar->field_order);
+    AKI_PACKET_WRITE_TYPE(packet, enum AVColorRange, codecpar->color_range);
+    AKI_PACKET_WRITE_TYPE(packet, enum AVColorPrimaries, codecpar->color_primaries);
+    AKI_PACKET_WRITE_TYPE(packet, enum AVColorTransferCharacteristic, codecpar->color_trc);
+    AKI_PACKET_WRITE_TYPE(packet, enum AVColorSpace, codecpar->color_space);
+    AKI_PACKET_WRITE_TYPE(packet, enum AVChromaLocation, codecpar->chroma_location);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->video_delay);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->sample_rate);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->block_align);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->frame_size);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->initial_padding);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->trailing_padding);
+    AKI_PACKET_WRITE_TYPE(packet, s32, codecpar->seek_preroll);
+    AKI_PACKET_WRITE_TYPE(packet, AVChannelLayout, codecpar->ch_layout);
+}
+
+void aki_packet_write_av_rational(struct aki_packet *packet, AVRational rational)
+{
+    AKI_PACKET_WRITE_TYPE(packet, AVRational, rational);
+}
+
+void aki_packet_write_av_codec_id(struct aki_packet *packet, enum AVCodecID codec_id)
+{
+    AKI_PACKET_WRITE_TYPE(packet, enum AVCodecID, codec_id);
+}
+
+void aki_packet_write_av_stream(struct aki_packet *packet, AVStream *stream)
+{
+    AKI_PACKET_WRITE_TYPE(packet, s32, stream->index);
+    aki_packet_write_av_codec_parameters(packet, stream->codecpar);
+    aki_packet_write_av_rational(packet, stream->time_base);
+    AKI_PACKET_WRITE_TYPE(packet, s64, stream->duration);
+    AKI_PACKET_WRITE_TYPE(packet, s64, stream->start_time);
+    AKI_PACKET_WRITE_TYPE(packet, s64, stream->nb_frames);
+    if (stream->avg_frame_rate.den == 0) {
+        // r_frame_rate.den == 0 handled on the client.
+        aki_packet_write_av_rational(packet, stream->r_frame_rate);
+    } else {
+        aki_packet_write_av_rational(packet, stream->avg_frame_rate);
+    }
+}
+
+void aki_packet_write_av_packet(struct aki_packet *packet, AVPacket *pkt)
+{
+    AKI_PACKET_WRITE_TYPE(packet, s64, pkt->pts);
+    AKI_PACKET_WRITE_TYPE(packet, s64, pkt->dts);
+    AKI_PACKET_WRITE_TYPE(packet, s32, pkt->size);
+    AKI_PACKET_WRITE_DATA(packet, pkt->data, pkt->size);
+    AKI_PACKET_WRITE_TYPE(packet, s32, pkt->stream_index);
+    AKI_PACKET_WRITE_TYPE(packet, s32, pkt->flags);
+    AKI_PACKET_WRITE_TYPE(packet, s32, pkt->side_data_elems);
+    for (s32 i = 0; i < pkt->side_data_elems; i++) {
+        AKI_PACKET_WRITE_TYPE(packet, size_t, pkt->side_data[i].size);
+        AKI_PACKET_WRITE_DATA(packet, pkt->side_data[i].data, pkt->side_data[i].size);
+        AKI_PACKET_WRITE_TYPE(packet, enum AVPacketSideDataType, pkt->side_data[i].type);
+    }
+    AKI_PACKET_WRITE_TYPE(packet, s64, pkt->duration);
+    AKI_PACKET_WRITE_TYPE(packet, s64, pkt->pos);
+    AKI_PACKET_WRITE_TYPE(packet, AVRational, pkt->time_base);
+}
+
+void aki_packet_read_av_codec_parameters(struct aki_packet *packet, AVCodecParameters *codecpar)
+{
+    AKI_PACKET_READ_TYPE(packet, enum AVMediaType, codecpar->codec_type);
+    AKI_PACKET_READ_TYPE(packet, enum AVCodecID, codecpar->codec_id);
+    AKI_PACKET_READ_TYPE(packet, u32, codecpar->codec_tag);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->extradata_size);
+    codecpar->extradata = av_malloc(codecpar->extradata_size);
+    u8 *extradata;
+    AKI_PACKET_READ_DATA(packet, codecpar->extradata_size, extradata);
+    al_memcpy(codecpar->extradata, extradata, codecpar->extradata_size);
+    AKI_PACKET_READ_TYPE(packet, s32,  codecpar->format);
+    AKI_PACKET_READ_TYPE(packet, s64, codecpar->bit_rate);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->bits_per_coded_sample);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->bits_per_raw_sample);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->profile);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->level);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->width);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->height);
+    AKI_PACKET_READ_TYPE(packet, AVRational, codecpar->sample_aspect_ratio);
+    AKI_PACKET_READ_TYPE(packet, enum AVFieldOrder, codecpar->field_order);
+    AKI_PACKET_READ_TYPE(packet, enum AVColorRange, codecpar->color_range);
+    AKI_PACKET_READ_TYPE(packet, enum AVColorPrimaries, codecpar->color_primaries);
+    AKI_PACKET_READ_TYPE(packet, enum AVColorTransferCharacteristic, codecpar->color_trc);
+    AKI_PACKET_READ_TYPE(packet, enum AVColorSpace, codecpar->color_space);
+    AKI_PACKET_READ_TYPE(packet, enum AVChromaLocation, codecpar->chroma_location);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->video_delay);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->sample_rate);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->block_align);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->frame_size);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->initial_padding);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->trailing_padding);
+    AKI_PACKET_READ_TYPE(packet, s32, codecpar->seek_preroll);
+    AKI_PACKET_READ_TYPE(packet, AVChannelLayout, codecpar->ch_layout);
+}
+
+AVRational aki_packet_read_av_rational(struct aki_packet *packet)
+{
+    AVRational r;
+    AKI_PACKET_READ_TYPE(packet, AVRational, r);
+    return r;
+}
+
+enum AVCodecID aki_packet_read_av_codec_id(struct aki_packet *packet)
+{
+    enum AVCodecID id;
+    AKI_PACKET_READ_TYPE(packet, enum AVCodecID, id);
+    return id;
+}
+
+AVStream *aki_packet_read_av_stream(AVFormatContext *format_context, const AVCodec *codec, struct aki_packet *packet)
+{
+    AVStream *stream = avformat_new_stream(format_context, codec);
+    AKI_PACKET_READ_TYPE(packet, s32, stream->index);
+    aki_packet_read_av_codec_parameters(packet, stream->codecpar);
+    stream->time_base = aki_packet_read_av_rational(packet);
+    AKI_PACKET_READ_TYPE(packet, s64, stream->duration);
+    AKI_PACKET_READ_TYPE(packet, s64, stream->start_time);
+    AKI_PACKET_READ_TYPE(packet, s64, stream->nb_frames);
+    stream->avg_frame_rate = aki_packet_read_av_rational(packet);
+    return stream;
+}
+
+AVPacket *aki_packet_read_av_packet(struct aki_packet *packet)
+{
+    AVPacket *pkt = av_packet_alloc();
+    AKI_PACKET_READ_TYPE(packet, s64, pkt->pts);
+    AKI_PACKET_READ_TYPE(packet, s64, pkt->dts);
+    AKI_PACKET_READ_TYPE(packet, s32, pkt->size);
+    AKI_PACKET_READ_DATA(packet, pkt->size, pkt->data);
+    AKI_PACKET_READ_TYPE(packet, s32, pkt->stream_index);
+    AKI_PACKET_READ_TYPE(packet, s32, pkt->flags);
+    s32 side_data_elems;
+    AKI_PACKET_READ_TYPE(packet, s32, side_data_elems);
+    pkt->side_data_elems = 0;
+    for (s32 i = 0; i < side_data_elems; i++) {
+        size_t size;
+        u8 *data;
+        enum AVPacketSideDataType type;
+        AKI_PACKET_READ_TYPE(packet, s32, size);
+        AKI_PACKET_READ_DATA(packet, size, data);
+        AKI_PACKET_READ_TYPE(packet, enum AVPacketSideDataType, type);
+        u8 *side_data = av_packet_new_side_data(pkt, type, size);
+        al_memcpy(side_data, data, size);
+    }
+    AKI_PACKET_READ_TYPE(packet, s64, pkt->duration);
+    AKI_PACKET_READ_TYPE(packet, s64, pkt->pos);
+    AKI_PACKET_READ_TYPE(packet, AVRational, pkt->time_base);
+    return pkt;
+}
diff --git a/src/codec/libav/packet_ext.h b/src/codec/libav/packet_ext.h
new file mode 100644
index 0000000..e2f7cd5
--- /dev/null
+++ b/src/codec/libav/packet_ext.h
@@ -0,0 +1,19 @@
+#pragma once
+
+#include <aki/common.h>
+
+#include <libavcodec/packet.h>
+#include <libavcodec/codec_par.h>
+#include <libavformat/avformat.h>
+
+void aki_packet_write_av_codec_parameters(struct aki_packet *packet, AVCodecParameters *codecpar);
+void aki_packet_write_av_rational(struct aki_packet *packet, AVRational rational);
+void aki_packet_write_av_codec_id(struct aki_packet *packet, enum AVCodecID codec_id);
+void aki_packet_write_av_stream(struct aki_packet *packet, AVStream *stream);
+void aki_packet_write_av_packet(struct aki_packet *packet, AVPacket *pkt);
+
+void aki_packet_read_av_codec_parameters(struct aki_packet *packet, AVCodecParameters *codecpar);
+AVRational aki_packet_read_av_rational(struct aki_packet *packet);
+enum AVCodecID aki_packet_read_av_codec_id(struct aki_packet *packet);
+AVStream *aki_packet_read_av_stream(AVFormatContext *format_context, const AVCodec *codec, struct aki_packet *packet);
+AVPacket *aki_packet_read_av_packet(struct aki_packet *packet);
diff --git a/src/codec/libav/resampler.c b/src/codec/libav/resampler.c
new file mode 100644
index 0000000..3dcb6d4
--- /dev/null
+++ b/src/codec/libav/resampler.c
@@ -0,0 +1,163 @@
+#include <libavutil/channel_layout.h>
+#include <libavutil/frame.h>
+#include <libavutil/opt.h>
+#include <libavutil/samplefmt.h>
+#include <libswresample/swresample.h>
+#include <al/lib.h>
+#include <al/log.h>
+
+#include "resampler.h"
+
+static bool in_req_match(struct camu_lav_resample_fmt *fmt)
+{
+    return (av_channel_layout_compare(&fmt->in_channel_layout, &fmt->req_channel_layout) == 0) &&
+      (fmt->in_format == fmt->req_format) && (fmt->in_sample_rate == fmt->req_sample_rate);
+}
+
+bool camu_lav_resampler_init(struct camu_lav_resampler *resamp, struct camu_lav_resample_fmt *fmt)
+{
+    al_memset(resamp, 0, sizeof(struct camu_lav_resampler));
+
+    fmt->resampler_needed = !in_req_match(fmt);
+
+    camu_lav_resample_fmt_copy(&resamp->fmt, fmt);
+
+    if (!fmt->resampler_needed) {
+        return true;
+    }
+
+    resamp->resample_context = swr_alloc();
+
+    SwrContext *ctx = resamp->resample_context;
+
+    if (!ctx) {
+        al_log_error("lav_resampler", "Failed to create resampler context.");
+        return false;
+    }
+
+    av_opt_set_chlayout(ctx,   "in_chlayout",    &fmt->in_channel_layout,  0);
+    av_opt_set_chlayout(ctx,   "out_chlayout",   &fmt->req_channel_layout, 0);
+    av_opt_set_int(ctx,        "in_sample_rate",  fmt->in_sample_rate,     0);
+    av_opt_set_int(ctx,        "out_sample_rate", fmt->req_sample_rate,    0);
+    av_opt_set_sample_fmt(ctx, "in_sample_fmt",   fmt->in_format,          0);
+    av_opt_set_sample_fmt(ctx, "out_sample_fmt",  fmt->req_format,         0);
+
+    // Slower but more accurate resampler.
+    av_opt_set_int(ctx, "resampler", SWR_ENGINE_SOXR, 0);
+
+    if (swr_init(ctx) != 0) {
+        swr_free(&ctx);
+        al_log_error("lav_resampler", "Failed to open resampler context.");
+        return false;
+    }
+
+    return true;
+}
+
+s32 camu_lav_resampler_sample_count(struct camu_lav_resampler *resamp, s32 in_samples)
+{
+    al_assert(resamp->fmt.resampler_needed);
+    return swr_get_out_samples(resamp->resample_context, in_samples);
+}
+
+static inline void free_sample_data(struct camu_lav_resampler *resamp)
+{
+    if (resamp->data[0]) av_freep(&resamp->data[0]);
+}
+
+static inline void alloc_sample_data(struct camu_lav_resampler *resamp, s32 sample_count)
+{
+    if (resamp->sample_count >= sample_count) return;
+    free_sample_data(resamp);
+    s32 nb_channels = resamp->fmt.req_channel_layout.nb_channels;
+    av_samples_alloc(resamp->data, NULL, nb_channels, sample_count, resamp->fmt.req_format, 0);
+    resamp->sample_count = sample_count;
+}
+
+s32 camu_lav_resampler_convert(struct camu_lav_resampler *resamp, const u8 **in_data, s32 in_samples)
+{
+    al_assert(resamp->fmt.resampler_needed);
+    s32 max_resample_count = in_samples ?
+      camu_lav_resampler_sample_count(resamp, in_samples) : resamp->sample_count;
+    alloc_sample_data(resamp, max_resample_count);
+    s32 resample_count = swr_convert(resamp->resample_context, resamp->data,
+        max_resample_count, in_data, in_samples);
+    if (resample_count < 0) {
+        al_log_error("lav_resampler", "Failed to convert samples (%i).", resample_count);
+    }
+    return resample_count;
+}
+
+u8 **camu_lav_resampler_get_data(struct camu_lav_resampler *resamp)
+{
+    return (u8 **)resamp->data;
+}
+
+void camu_lav_resampler_close(struct camu_lav_resampler *resamp)
+{
+    al_assert(resamp->fmt.resampler_needed);
+    if (resamp->resample_context) {
+        swr_close(resamp->resample_context);
+        swr_free(&resamp->resample_context);
+    }
+    free_sample_data(resamp);
+}
+
+void camu_lav_resample_fmt_copy(struct camu_lav_resample_fmt *dest, struct camu_lav_resample_fmt *src)
+{
+    dest->in_format = src->in_format;
+    av_channel_layout_copy(&dest->in_channel_layout, &src->in_channel_layout);
+    dest->in_sample_rate = src->in_sample_rate;
+    dest->req_format = src->req_format;
+    av_channel_layout_copy(&dest->req_channel_layout, &src->req_channel_layout);
+    dest->req_sample_rate = src->req_sample_rate;
+    dest->resampler_needed = src->resampler_needed;
+}
+
+size_t camu_lav_resample_fmt_bytes_per_sample(struct camu_lav_resample_fmt *fmt)
+{
+    return (size_t)av_get_bytes_per_sample(fmt->req_format);
+}
+
+size_t camu_lav_resample_fmt_samples_to_bytes(struct camu_lav_resample_fmt *fmt, size_t samples)
+{
+    s32 size, nb_channels = fmt->req_channel_layout.nb_channels;
+    if (av_samples_get_buffer_size(&size, nb_channels, (s32)samples, fmt->req_format, 1) < 0) {
+        return 0;
+    }
+    return (size_t)size;
+}
+
+size_t camu_lav_resample_fmt_samples_to_usec(struct camu_lav_resample_fmt *fmt, size_t samples)
+{
+    return samples / (fmt->req_sample_rate / 1000000.0);
+}
+
+f64 camu_lav_resample_fmt_samples_to_sec(struct camu_lav_resample_fmt *fmt, size_t samples)
+{
+    return samples / (f64)fmt->req_sample_rate;
+}
+
+size_t camu_lav_resample_fmt_usec_to_bytes(struct camu_lav_resample_fmt *fmt, size_t usec)
+{
+    size_t samples = usec * (fmt->req_sample_rate / 1000000.0);
+    return samples * (camu_lav_resample_fmt_bytes_per_sample(fmt) * fmt->req_channel_layout.nb_channels);
+}
+
+size_t camu_lav_resample_fmt_sec_to_bytes(struct camu_lav_resample_fmt *fmt, f64 sec)
+{
+    size_t samples = sec * fmt->req_sample_rate;
+    return samples * (camu_lav_resample_fmt_bytes_per_sample(fmt) * fmt->req_channel_layout.nb_channels);
+}
+
+size_t camu_lav_resample_fmt_bytes_to_usec(struct camu_lav_resample_fmt *fmt, size_t bytes)
+{
+    size_t samples = bytes / (camu_lav_resample_fmt_bytes_per_sample(fmt) * fmt->req_channel_layout.nb_channels);
+    return samples / (fmt->req_sample_rate / 1000000.0);
+}
+
+f64 camu_lav_resample_fmt_bytes_to_sec(struct camu_lav_resample_fmt *fmt, size_t bytes)
+{
+    size_t samples = bytes / (camu_lav_resample_fmt_bytes_per_sample(fmt) * fmt->req_channel_layout.nb_channels);
+    return samples / ((f64)fmt->req_sample_rate);
+}
diff --git a/src/codec/libav/resampler.h b/src/codec/libav/resampler.h
new file mode 100644
index 0000000..703d9a7
--- /dev/null
+++ b/src/codec/libav/resampler.h
@@ -0,0 +1,40 @@
+#pragma once
+
+#include <libswresample/swresample.h>
+#include <libavutil/opt.h>
+#include <al/types.h>
+
+struct camu_lav_resample_fmt {
+    s32 in_channel_count;
+    s32 req_channel_count;
+    AVChannelLayout in_channel_layout;
+    AVChannelLayout req_channel_layout;
+    s32 in_sample_rate;
+    s32 req_sample_rate;
+    enum AVSampleFormat in_format;
+    enum AVSampleFormat req_format;
+    bool resampler_needed;
+};
+
+struct camu_lav_resampler {
+    SwrContext *resample_context;
+    struct camu_lav_resample_fmt fmt;
+    u8 *data[AV_NUM_DATA_POINTERS];
+    s32 sample_count;
+};
+
+bool camu_lav_resampler_init(struct camu_lav_resampler *resamp, struct camu_lav_resample_fmt *fmt);
+s32 camu_lav_resampler_sample_count(struct camu_lav_resampler *resamp, s32 in_samples);
+s32 camu_lav_resampler_convert(struct camu_lav_resampler *resamp, const u8 **in_data, s32 in_samples);
+u8 **camu_lav_resampler_get_data(struct camu_lav_resampler *resamp);
+void camu_lav_resampler_close(struct camu_lav_resampler *resamp);
+
+void camu_lav_resample_fmt_copy(struct camu_lav_resample_fmt *dest, struct camu_lav_resample_fmt *src);
+size_t camu_lav_resample_fmt_bytes_per_sample(struct camu_lav_resample_fmt *fmt);
+size_t camu_lav_resample_fmt_samples_to_bytes(struct camu_lav_resample_fmt *fmt, size_t samples);
+size_t camu_lav_resample_fmt_samples_to_usec(struct camu_lav_resample_fmt *fmt, size_t samples);
+f64 camu_lav_resample_fmt_samples_to_sec(struct camu_lav_resample_fmt *fmt, size_t samples);
+size_t camu_lav_resample_fmt_usec_to_bytes(struct camu_lav_resample_fmt *fmt, size_t usec);
+size_t camu_lav_resample_fmt_sec_to_bytes(struct camu_lav_resample_fmt *fmt, f64 sec);
+size_t camu_lav_resample_fmt_bytes_to_usec(struct camu_lav_resample_fmt *fmt, size_t bytes);
+f64 camu_lav_resample_fmt_bytes_to_sec(struct camu_lav_resample_fmt *fmt, size_t bytes);
diff --git a/src/codec/libav/scaler.c b/src/codec/libav/scaler.c
new file mode 100644
index 0000000..6146c69
--- /dev/null
+++ b/src/codec/libav/scaler.c
@@ -0,0 +1,79 @@
+#include <al/lib.h>
+#include <al/log.h>
+#include <libavutil/imgutils.h>
+#include <libavcodec/avcodec.h>
+
+#include "scaler.h"
+
+static AVFrame *alloc_picture(enum AVPixelFormat pix_fmt, s32 width, s32 height)
+{
+    AVFrame *frame = av_frame_alloc();
+    if (!frame) return NULL;
+    s32 size = av_image_get_buffer_size(pix_fmt, width, height, 1);
+    u8 *buffer = (u8 *)av_malloc(size);
+    if (!buffer) {
+        av_free(frame);
+        return NULL;
+    }
+    s32 ret = av_image_fill_arrays(frame->data, frame->linesize, buffer, pix_fmt, width, height, 1);
+    if (ret < 0) {
+        al_log_error("lav_scaler", "Failed to allocate picture buffer (%s).", av_err2str(ret));
+        av_free(frame);
+        return NULL;
+    }
+    frame->width = width;
+    frame->height = height;
+    frame->format = pix_fmt;
+    return frame;
+}
+
+bool camu_lav_scaler_init(struct camu_lav_scaler *scale, struct camu_lav_scale_fmt *fmt)
+{
+    al_memset(scale, 0, sizeof(struct camu_lav_scaler));
+
+    /*
+    if (fmt->in_width == fmt->req_width && fmt->in_height == fmt->req_height &&
+        fmt->in_format == fmt->req_format) {
+        fmt->scaler_needed = false;
+        return true;
+    }
+    */
+
+    fmt->scaler_needed = true;
+
+    scale->scaler_context = sws_getContext(fmt->in_width, fmt->in_height, fmt->in_format,
+        fmt->req_width, fmt->req_height, fmt->req_format,
+        SWS_BITEXACT, NULL, NULL, NULL);
+
+    if (!scale->scaler_context) {
+        al_log_error("lav_scaler", "Failed to create scaler context.");
+        return false;
+    }
+
+    scale->frame = alloc_picture(fmt->req_format, fmt->req_width, fmt->req_height);
+
+    if (!scale->frame) {
+        al_log_error("lav_scaler", "Failed to allocate frame.");
+        return false;
+    }
+
+    scale->fmt = *fmt;
+
+    return true;
+}
+
+bool camu_lav_scaler_scale(struct camu_lav_scaler *scale, const u8 **in_slice, s32 *in_strides)
+{
+    s32 ret = sws_scale(scale->scaler_context, in_slice, in_strides, 0,
+        scale->fmt.in_height, scale->frame->data, scale->frame->linesize);
+    if (ret != scale->fmt.req_height) {
+        al_log_error("lav_scaler", "Failed to scale frame (%s).", av_err2str(ret));
+        return false;
+    }
+    return true;
+}
+
+void camu_lav_scaler_close(struct camu_lav_scaler *scale)
+{
+    sws_freeContext(scale->scaler_context);
+}
diff --git a/src/codec/libav/scaler.h b/src/codec/libav/scaler.h
new file mode 100644
index 0000000..4d54085
--- /dev/null
+++ b/src/codec/libav/scaler.h
@@ -0,0 +1,24 @@
+#pragma once
+
+#include <libswscale/swscale.h>
+#include <al/types.h>
+
+struct camu_lav_scale_fmt {
+    s32 in_width;
+    s32 in_height;
+    enum AVPixelFormat in_format;
+    s32 req_width;
+    s32 req_height;
+    enum AVPixelFormat req_format;
+    bool scaler_needed;
+};
+
+struct camu_lav_scaler {
+    struct SwsContext *scaler_context;
+    struct camu_lav_scale_fmt fmt;
+    AVFrame *frame;
+};
+
+bool camu_lav_scaler_init(struct camu_lav_scaler *scale, struct camu_lav_scale_fmt *fmt);
+bool camu_lav_scaler_scale(struct camu_lav_scaler *scale, const u8 **in_slice, s32 *in_strides);
+void camu_lav_scaler_close(struct camu_lav_scaler *scale);