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;
51 float *samples_flt[2];
53 AVFloatDSPContext fdsp;
57 static int realloc_buffer(LAMEContext *s)
59 if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) {
61 int new_size = s->buffer_index + 2 * BUFFER_SIZE;
63 av_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size,
65 tmp = av_realloc(s->buffer, new_size);
68 s->buffer_size = s->buffer_index = 0;
69 return AVERROR(ENOMEM);
72 s->buffer_size = new_size;
77 static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
79 LAMEContext *s = avctx->priv_data;
81 av_freep(&s->samples_flt[0]);
82 av_freep(&s->samples_flt[1]);
85 ff_af_queue_close(&s->afq);
91 static av_cold int mp3lame_encode_init(AVCodecContext *avctx)
93 LAMEContext *s = avctx->priv_data;
98 /* initialize LAME and get defaults */
99 if ((s->gfp = lame_init()) == NULL)
100 return AVERROR(ENOMEM);
102 lame_set_num_channels(s->gfp, avctx->channels);
103 lame_set_mode(s->gfp, avctx->channels > 1 ? JOINT_STEREO : MONO);
106 lame_set_in_samplerate (s->gfp, avctx->sample_rate);
107 lame_set_out_samplerate(s->gfp, avctx->sample_rate);
109 /* algorithmic quality */
110 if (avctx->compression_level == FF_COMPRESSION_DEFAULT)
111 lame_set_quality(s->gfp, 5);
113 lame_set_quality(s->gfp, avctx->compression_level);
116 if (avctx->flags & CODEC_FLAG_QSCALE) {
117 lame_set_VBR(s->gfp, vbr_default);
118 lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
121 lame_set_brate(s->gfp, avctx->bit_rate / 1000);
124 /* do not get a Xing VBR header frame from LAME */
125 lame_set_bWriteVbrTag(s->gfp,0);
127 /* bit reservoir usage */
128 lame_set_disable_reservoir(s->gfp, !s->reservoir);
130 /* set specified parameters */
131 if (lame_init_params(s->gfp) < 0) {
136 /* get encoder delay */
137 avctx->delay = lame_get_encoder_delay(s->gfp) + 528 + 1;
138 ff_af_queue_init(avctx, &s->afq);
140 avctx->frame_size = lame_get_framesize(s->gfp);
142 /* allocate float sample buffers */
143 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
145 for (ch = 0; ch < avctx->channels; ch++) {
146 s->samples_flt[ch] = av_malloc(avctx->frame_size *
147 sizeof(*s->samples_flt[ch]));
148 if (!s->samples_flt[ch]) {
149 ret = AVERROR(ENOMEM);
155 ret = realloc_buffer(s);
159 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
163 mp3lame_encode_close(avctx);
167 #define ENCODE_BUFFER(func, buf_type, buf_name) do { \
168 lame_result = func(s->gfp, \
169 (const buf_type *)buf_name[0], \
170 (const buf_type *)buf_name[1], frame->nb_samples, \
171 s->buffer + s->buffer_index, \
172 s->buffer_size - s->buffer_index); \
175 static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
176 const AVFrame *frame, int *got_packet_ptr)
178 LAMEContext *s = avctx->priv_data;
184 switch (avctx->sample_fmt) {
185 case AV_SAMPLE_FMT_S16P:
186 ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data);
188 case AV_SAMPLE_FMT_S32P:
189 ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data);
191 case AV_SAMPLE_FMT_FLTP:
192 if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) {
193 av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n");
194 return AVERROR(EINVAL);
196 for (ch = 0; ch < avctx->channels; ch++) {
197 s->fdsp.vector_fmul_scalar(s->samples_flt[ch],
198 (const float *)frame->data[ch],
200 FFALIGN(frame->nb_samples, 8));
202 ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt);
208 lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
209 s->buffer_size - s->buffer_index);
211 if (lame_result < 0) {
212 if (lame_result == -1) {
213 av_log(avctx, AV_LOG_ERROR,
214 "lame: output buffer too small (buffer index: %d, free bytes: %d)\n",
215 s->buffer_index, s->buffer_size - s->buffer_index);
219 s->buffer_index += lame_result;
220 ret = realloc_buffer(s);
222 av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n");
226 /* add current frame to the queue */
228 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
232 /* Move 1 frame from the LAME buffer to the output packet, if available.
233 We have to parse the first frame header in the output buffer to
234 determine the frame size. */
235 if (s->buffer_index < 4)
237 if (avpriv_mpegaudio_decode_header(&hdr, AV_RB32(s->buffer))) {
238 av_log(avctx, AV_LOG_ERROR, "free format output not supported\n");
241 len = hdr.frame_size;
242 av_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
244 if (len <= s->buffer_index) {
245 if ((ret = ff_alloc_packet(avpkt, len))) {
246 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
249 memcpy(avpkt->data, s->buffer, len);
250 s->buffer_index -= len;
251 memmove(s->buffer, s->buffer + len, s->buffer_index);
253 /* Get the next frame pts/duration */
254 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
263 #define OFFSET(x) offsetof(LAMEContext, x)
264 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
265 static const AVOption options[] = {
266 { "reservoir", "Use bit reservoir.", OFFSET(reservoir), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
270 static const AVClass libmp3lame_class = {
271 .class_name = "libmp3lame encoder",
272 .item_name = av_default_item_name,
274 .version = LIBAVUTIL_VERSION_INT,
277 static const AVCodecDefault libmp3lame_defaults[] = {
282 static const int libmp3lame_sample_rates[] = {
283 44100, 48000, 32000, 22050, 24000, 16000, 11025, 12000, 8000, 0
286 AVCodec ff_libmp3lame_encoder = {
287 .name = "libmp3lame",
288 .type = AVMEDIA_TYPE_AUDIO,
289 .id = AV_CODEC_ID_MP3,
290 .priv_data_size = sizeof(LAMEContext),
291 .init = mp3lame_encode_init,
292 .encode2 = mp3lame_encode_frame,
293 .close = mp3lame_encode_close,
294 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME,
295 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P,
298 AV_SAMPLE_FMT_NONE },
299 .supported_samplerates = libmp3lame_sample_rates,
300 .channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
303 .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
304 .priv_class = &libmp3lame_class,
305 .defaults = libmp3lame_defaults,