2 * Interface to libmp3lame for mp3 encoding
3 * Copyright (c) 2002 Lennert Buytenhek <buytenh@gnu.org>
5 * This file is part of Libav.
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Interface to libmp3lame for mp3 encoding.
27 #include <lame/lame.h>
29 #include "libavutil/channel_layout.h"
30 #include "libavutil/common.h"
31 #include "libavutil/float_dsp.h"
32 #include "libavutil/intreadwrite.h"
33 #include "libavutil/log.h"
34 #include "libavutil/opt.h"
36 #include "audio_frame_queue.h"
38 #include "mpegaudio.h"
39 #include "mpegaudiodecheader.h"
41 #define BUFFER_SIZE (7200 + 2 * MPA_FRAME_SIZE + MPA_FRAME_SIZE / 4)
43 typedef struct LAMEContext {
45 AVCodecContext *avctx;
46 lame_global_flags *gfp;
53 float *samples_flt[2];
55 AVFloatDSPContext fdsp;
59 static int realloc_buffer(LAMEContext *s)
61 if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) {
62 int new_size = s->buffer_index + 2 * BUFFER_SIZE, err;
64 ff_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size,
66 if ((err = av_reallocp(&s->buffer, new_size)) < 0) {
67 s->buffer_size = s->buffer_index = 0;
70 s->buffer_size = new_size;
75 static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
77 LAMEContext *s = avctx->priv_data;
79 av_freep(&s->samples_flt[0]);
80 av_freep(&s->samples_flt[1]);
83 ff_af_queue_close(&s->afq);
89 static av_cold int mp3lame_encode_init(AVCodecContext *avctx)
91 LAMEContext *s = avctx->priv_data;
96 /* initialize LAME and get defaults */
97 if (!(s->gfp = lame_init()))
98 return AVERROR(ENOMEM);
100 lame_set_num_channels(s->gfp, avctx->channels);
101 lame_set_mode(s->gfp, avctx->channels > 1 ? s->joint_stereo ? JOINT_STEREO : STEREO : MONO);
104 lame_set_in_samplerate (s->gfp, avctx->sample_rate);
105 lame_set_out_samplerate(s->gfp, avctx->sample_rate);
107 /* algorithmic quality */
108 if (avctx->compression_level == FF_COMPRESSION_DEFAULT)
109 lame_set_quality(s->gfp, 5);
111 lame_set_quality(s->gfp, avctx->compression_level);
114 if (avctx->flags & AV_CODEC_FLAG_QSCALE) { // VBR
115 lame_set_VBR(s->gfp, vbr_default);
116 lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
118 if (avctx->bit_rate) {
120 lame_set_VBR(s->gfp, vbr_abr);
121 lame_set_VBR_mean_bitrate_kbps(s->gfp, avctx->bit_rate / 1000);
123 lame_set_brate(s->gfp, avctx->bit_rate / 1000);
127 /* do not get a Xing VBR header frame from LAME */
128 lame_set_bWriteVbrTag(s->gfp,0);
130 /* bit reservoir usage */
131 lame_set_disable_reservoir(s->gfp, !s->reservoir);
133 /* set specified parameters */
134 if (lame_init_params(s->gfp) < 0) {
139 /* get encoder delay */
140 avctx->initial_padding = lame_get_encoder_delay(s->gfp) + 528 + 1;
141 ff_af_queue_init(avctx, &s->afq);
143 avctx->frame_size = lame_get_framesize(s->gfp);
145 /* allocate float sample buffers */
146 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
148 for (ch = 0; ch < avctx->channels; ch++) {
149 s->samples_flt[ch] = av_malloc(avctx->frame_size *
150 sizeof(*s->samples_flt[ch]));
151 if (!s->samples_flt[ch]) {
152 ret = AVERROR(ENOMEM);
158 ret = realloc_buffer(s);
162 avpriv_float_dsp_init(&s->fdsp, avctx->flags & AV_CODEC_FLAG_BITEXACT);
166 mp3lame_encode_close(avctx);
170 #define ENCODE_BUFFER(func, buf_type, buf_name) do { \
171 lame_result = func(s->gfp, \
172 (const buf_type *)buf_name[0], \
173 (const buf_type *)buf_name[1], frame->nb_samples, \
174 s->buffer + s->buffer_index, \
175 s->buffer_size - s->buffer_index); \
178 static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
179 const AVFrame *frame, int *got_packet_ptr)
181 LAMEContext *s = avctx->priv_data;
188 switch (avctx->sample_fmt) {
189 case AV_SAMPLE_FMT_S16P:
190 ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data);
192 case AV_SAMPLE_FMT_S32P:
193 ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data);
195 case AV_SAMPLE_FMT_FLTP:
196 if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) {
197 av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n");
198 return AVERROR(EINVAL);
200 for (ch = 0; ch < avctx->channels; ch++) {
201 s->fdsp.vector_fmul_scalar(s->samples_flt[ch],
202 (const float *)frame->data[ch],
204 FFALIGN(frame->nb_samples, 8));
206 ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt);
212 lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
213 s->buffer_size - s->buffer_index);
215 if (lame_result < 0) {
216 if (lame_result == -1) {
217 av_log(avctx, AV_LOG_ERROR,
218 "lame: output buffer too small (buffer index: %d, free bytes: %d)\n",
219 s->buffer_index, s->buffer_size - s->buffer_index);
223 s->buffer_index += lame_result;
224 ret = realloc_buffer(s);
226 av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n");
230 /* add current frame to the queue */
232 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
236 /* Move 1 frame from the LAME buffer to the output packet, if available.
237 We have to parse the first frame header in the output buffer to
238 determine the frame size. */
239 if (s->buffer_index < 4)
241 h = AV_RB32(s->buffer);
242 if (ff_mpa_check_header(h) < 0) {
243 av_log(avctx, AV_LOG_ERROR, "Invalid mp3 header at start of buffer\n");
246 if (avpriv_mpegaudio_decode_header(&hdr, h)) {
247 av_log(avctx, AV_LOG_ERROR, "free format output not supported\n");
250 len = hdr.frame_size;
251 ff_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
253 if (len <= s->buffer_index) {
254 if ((ret = ff_alloc_packet(avpkt, len))) {
255 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
258 memcpy(avpkt->data, s->buffer, len);
259 s->buffer_index -= len;
260 memmove(s->buffer, s->buffer + len, s->buffer_index);
262 /* Get the next frame pts/duration */
263 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
272 #define OFFSET(x) offsetof(LAMEContext, x)
273 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
274 static const AVOption options[] = {
275 { "reservoir", "Use bit reservoir.", OFFSET(reservoir), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
276 { "joint_stereo", "Use joint stereo.", OFFSET(joint_stereo), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
277 { "abr", "Use ABR", OFFSET(abr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE },
281 static const AVClass libmp3lame_class = {
282 .class_name = "libmp3lame encoder",
283 .item_name = av_default_item_name,
285 .version = LIBAVUTIL_VERSION_INT,
288 static const AVCodecDefault libmp3lame_defaults[] = {
293 static const int libmp3lame_sample_rates[] = {
294 44100, 48000, 32000, 22050, 24000, 16000, 11025, 12000, 8000, 0
297 AVCodec ff_libmp3lame_encoder = {
298 .name = "libmp3lame",
299 .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
300 .type = AVMEDIA_TYPE_AUDIO,
301 .id = AV_CODEC_ID_MP3,
302 .priv_data_size = sizeof(LAMEContext),
303 .init = mp3lame_encode_init,
304 .encode2 = mp3lame_encode_frame,
305 .close = mp3lame_encode_close,
306 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_SMALL_LAST_FRAME,
307 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P,
310 AV_SAMPLE_FMT_NONE },
311 .supported_samplerates = libmp3lame_sample_rates,
312 .channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
315 .priv_class = &libmp3lame_class,
316 .defaults = libmp3lame_defaults,