2 * Interface to libmp3lame for mp3 encoding
3 * Copyright (c) 2002 Lennert Buytenhek <buytenh@gnu.org>
5 * This file is part of FFmpeg.
7 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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+1000) // FIXME: Buffer size to small? Adding 1000 to make up for it.
43 typedef struct LAMEContext {
45 AVCodecContext *avctx;
46 lame_global_flags *gfp;
52 float *samples_flt[2];
54 AVFloatDSPContext fdsp;
58 static int realloc_buffer(LAMEContext *s)
60 if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) {
62 int new_size = s->buffer_index + 2 * BUFFER_SIZE;
64 av_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size,
66 tmp = av_realloc(s->buffer, new_size);
69 s->buffer_size = s->buffer_index = 0;
70 return AVERROR(ENOMEM);
73 s->buffer_size = new_size;
78 static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
80 LAMEContext *s = avctx->priv_data;
82 av_freep(&s->samples_flt[0]);
83 av_freep(&s->samples_flt[1]);
86 ff_af_queue_close(&s->afq);
92 static av_cold int mp3lame_encode_init(AVCodecContext *avctx)
94 LAMEContext *s = avctx->priv_data;
99 /* initialize LAME and get defaults */
100 if ((s->gfp = lame_init()) == NULL)
101 return AVERROR(ENOMEM);
104 lame_set_num_channels(s->gfp, avctx->channels);
105 lame_set_mode(s->gfp, avctx->channels > 1 ? s->joint_stereo ? JOINT_STEREO : STEREO : MONO);
108 lame_set_in_samplerate (s->gfp, avctx->sample_rate);
109 lame_set_out_samplerate(s->gfp, avctx->sample_rate);
111 /* algorithmic quality */
112 if (avctx->compression_level == FF_COMPRESSION_DEFAULT)
113 lame_set_quality(s->gfp, 5);
115 lame_set_quality(s->gfp, avctx->compression_level);
118 if (avctx->flags & CODEC_FLAG_QSCALE) {
119 lame_set_VBR(s->gfp, vbr_default);
120 lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
123 lame_set_brate(s->gfp, avctx->bit_rate / 1000);
126 /* do not get a Xing VBR header frame from LAME */
127 lame_set_bWriteVbrTag(s->gfp,0);
129 /* bit reservoir usage */
130 lame_set_disable_reservoir(s->gfp, !s->reservoir);
132 /* set specified parameters */
133 if (lame_init_params(s->gfp) < 0) {
138 /* get encoder delay */
139 avctx->delay = lame_get_encoder_delay(s->gfp) + 528 + 1;
140 ff_af_queue_init(avctx, &s->afq);
142 avctx->frame_size = lame_get_framesize(s->gfp);
144 /* allocate float sample buffers */
145 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
147 for (ch = 0; ch < avctx->channels; ch++) {
148 s->samples_flt[ch] = av_malloc(avctx->frame_size *
149 sizeof(*s->samples_flt[ch]));
150 if (!s->samples_flt[ch]) {
151 ret = AVERROR(ENOMEM);
157 ret = realloc_buffer(s);
161 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
165 mp3lame_encode_close(avctx);
169 #define ENCODE_BUFFER(func, buf_type, buf_name) do { \
170 lame_result = func(s->gfp, \
171 (const buf_type *)buf_name[0], \
172 (const buf_type *)buf_name[1], frame->nb_samples, \
173 s->buffer + s->buffer_index, \
174 s->buffer_size - s->buffer_index); \
177 static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
178 const AVFrame *frame, int *got_packet_ptr)
180 LAMEContext *s = avctx->priv_data;
186 switch (avctx->sample_fmt) {
187 case AV_SAMPLE_FMT_S16P:
188 ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data);
190 case AV_SAMPLE_FMT_S32P:
191 ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data);
193 case AV_SAMPLE_FMT_FLTP:
194 if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) {
195 av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n");
196 return AVERROR(EINVAL);
198 for (ch = 0; ch < avctx->channels; ch++) {
199 s->fdsp.vector_fmul_scalar(s->samples_flt[ch],
200 (const float *)frame->data[ch],
202 FFALIGN(frame->nb_samples, 8));
204 ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt);
210 lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
211 s->buffer_size - s->buffer_index);
213 if (lame_result < 0) {
214 if (lame_result == -1) {
215 av_log(avctx, AV_LOG_ERROR,
216 "lame: output buffer too small (buffer index: %d, free bytes: %d)\n",
217 s->buffer_index, s->buffer_size - s->buffer_index);
221 s->buffer_index += lame_result;
222 ret = realloc_buffer(s);
224 av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n");
228 /* add current frame to the queue */
230 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
234 /* Move 1 frame from the LAME buffer to the output packet, if available.
235 We have to parse the first frame header in the output buffer to
236 determine the frame size. */
237 if (s->buffer_index < 4)
239 if (avpriv_mpegaudio_decode_header(&hdr, AV_RB32(s->buffer))) {
240 av_log(avctx, AV_LOG_ERROR, "free format output not supported\n");
243 len = hdr.frame_size;
244 av_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
246 if (len <= s->buffer_index) {
247 if ((ret = ff_alloc_packet2(avctx, avpkt, len)) < 0)
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 },
267 { "joint_stereo", "Use joint stereo.", OFFSET(joint_stereo), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
271 static const AVClass libmp3lame_class = {
272 .class_name = "libmp3lame encoder",
273 .item_name = av_default_item_name,
275 .version = LIBAVUTIL_VERSION_INT,
278 static const AVCodecDefault libmp3lame_defaults[] = {
283 static const int libmp3lame_sample_rates[] = {
284 44100, 48000, 32000, 22050, 24000, 16000, 11025, 12000, 8000, 0
287 AVCodec ff_libmp3lame_encoder = {
288 .name = "libmp3lame",
289 .type = AVMEDIA_TYPE_AUDIO,
290 .id = AV_CODEC_ID_MP3,
291 .priv_data_size = sizeof(LAMEContext),
292 .init = mp3lame_encode_init,
293 .encode2 = mp3lame_encode_frame,
294 .close = mp3lame_encode_close,
295 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME,
296 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P,
299 AV_SAMPLE_FMT_NONE },
300 .supported_samplerates = libmp3lame_sample_rates,
301 .channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
304 .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
305 .priv_class = &libmp3lame_class,
306 .defaults = libmp3lame_defaults,