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/intreadwrite.h"
30 #include "libavutil/log.h"
31 #include "libavutil/opt.h"
33 #include "audio_frame_queue.h"
35 #include "mpegaudio.h"
36 #include "mpegaudiodecheader.h"
38 #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.
40 typedef struct LAMEContext {
42 AVCodecContext *avctx;
43 lame_global_flags *gfp;
44 uint8_t buffer[BUFFER_SIZE];
47 void *planar_samples[2];
52 static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
54 LAMEContext *s = avctx->priv_data;
56 #if FF_API_OLD_ENCODE_AUDIO
57 av_freep(&avctx->coded_frame);
59 av_freep(&s->planar_samples[0]);
60 av_freep(&s->planar_samples[1]);
62 ff_af_queue_close(&s->afq);
68 static av_cold int mp3lame_encode_init(AVCodecContext *avctx)
70 LAMEContext *s = avctx->priv_data;
75 /* initialize LAME and get defaults */
76 if ((s->gfp = lame_init()) == NULL)
77 return AVERROR(ENOMEM);
80 if (avctx->channels > 2) {
81 av_log(avctx, AV_LOG_ERROR,
82 "Invalid number of channels %d, must be <= 2\n", avctx->channels);
83 ret = AVERROR(EINVAL);
86 lame_set_num_channels(s->gfp, avctx->channels);
87 lame_set_mode(s->gfp, avctx->channels > 1 ? JOINT_STEREO : MONO);
90 lame_set_in_samplerate (s->gfp, avctx->sample_rate);
91 lame_set_out_samplerate(s->gfp, avctx->sample_rate);
93 /* algorithmic quality */
94 if (avctx->compression_level == FF_COMPRESSION_DEFAULT)
95 lame_set_quality(s->gfp, 5);
97 lame_set_quality(s->gfp, avctx->compression_level);
100 if (avctx->flags & CODEC_FLAG_QSCALE) {
101 lame_set_VBR(s->gfp, vbr_default);
102 lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
105 lame_set_brate(s->gfp, avctx->bit_rate / 1000);
108 /* do not get a Xing VBR header frame from LAME */
109 lame_set_bWriteVbrTag(s->gfp,0);
111 /* bit reservoir usage */
112 lame_set_disable_reservoir(s->gfp, !s->reservoir);
114 /* set specified parameters */
115 if (lame_init_params(s->gfp) < 0) {
120 /* get encoder delay */
121 avctx->delay = lame_get_encoder_delay(s->gfp) + 528 + 1;
122 ff_af_queue_init(avctx, &s->afq);
124 avctx->frame_size = lame_get_framesize(s->gfp);
126 #if FF_API_OLD_ENCODE_AUDIO
127 avctx->coded_frame = avcodec_alloc_frame();
128 if (!avctx->coded_frame) {
129 ret = AVERROR(ENOMEM);
135 if (avctx->sample_fmt == AV_SAMPLE_FMT_S32 ||
136 avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
138 for (ch = 0; ch < avctx->channels; ch++) {
139 s->planar_samples[ch] = av_malloc(avctx->frame_size *
140 av_get_bytes_per_sample(avctx->sample_fmt));
141 if (!s->planar_samples[ch]) {
142 ret = AVERROR(ENOMEM);
150 mp3lame_encode_close(avctx);
154 #define DEINTERLEAVE(type, scale) do { \
156 for (ch = 0; ch < s->avctx->channels; ch++) { \
157 const type *input = samples; \
158 type *output = s->planar_samples[ch]; \
160 for (i = 0; i < nb_samples; i++) { \
161 output[i] = *input * scale; \
162 input += s->avctx->channels; \
167 static int encode_frame_int16(LAMEContext *s, void *samples, int nb_samples)
169 if (s->avctx->channels > 1) {
170 return lame_encode_buffer_interleaved(s->gfp, samples,
172 s->buffer + s->buffer_index,
173 BUFFER_SIZE - s->buffer_index);
175 return lame_encode_buffer(s->gfp, samples, NULL, nb_samples,
176 s->buffer + s->buffer_index,
177 BUFFER_SIZE - s->buffer_index);
181 static int encode_frame_int32(LAMEContext *s, void *samples, int nb_samples)
183 DEINTERLEAVE(int32_t, 1);
185 return lame_encode_buffer_int(s->gfp,
186 s->planar_samples[0], s->planar_samples[1],
188 s->buffer + s->buffer_index,
189 BUFFER_SIZE - s->buffer_index);
192 static int encode_frame_float(LAMEContext *s, void *samples, int nb_samples)
194 DEINTERLEAVE(float, 32768.0f);
196 return lame_encode_buffer_float(s->gfp,
197 s->planar_samples[0], s->planar_samples[1],
199 s->buffer + s->buffer_index,
200 BUFFER_SIZE - s->buffer_index);
203 static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
204 const AVFrame *frame, int *got_packet_ptr)
206 LAMEContext *s = avctx->priv_data;
212 switch (avctx->sample_fmt) {
213 case AV_SAMPLE_FMT_S16:
214 lame_result = encode_frame_int16(s, frame->data[0], frame->nb_samples);
216 case AV_SAMPLE_FMT_S32:
217 lame_result = encode_frame_int32(s, frame->data[0], frame->nb_samples);
219 case AV_SAMPLE_FMT_FLT:
220 lame_result = encode_frame_float(s, frame->data[0], frame->nb_samples);
226 lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
227 BUFFER_SIZE - s->buffer_index);
229 if (lame_result < 0) {
230 if (lame_result == -1) {
231 av_log(avctx, AV_LOG_ERROR,
232 "lame: output buffer too small (buffer index: %d, free bytes: %d)\n",
233 s->buffer_index, BUFFER_SIZE - s->buffer_index);
237 s->buffer_index += lame_result;
239 /* add current frame to the queue */
241 if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
245 /* Move 1 frame from the LAME buffer to the output packet, if available.
246 We have to parse the first frame header in the output buffer to
247 determine the frame size. */
248 if (s->buffer_index < 4)
250 if (avpriv_mpegaudio_decode_header(&hdr, AV_RB32(s->buffer))) {
251 av_log(avctx, AV_LOG_ERROR, "free format output not supported\n");
254 len = hdr.frame_size;
255 av_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
257 if (len <= s->buffer_index) {
258 if ((ret = ff_alloc_packet2(avctx, avpkt, len)))
260 memcpy(avpkt->data, s->buffer, len);
261 s->buffer_index -= len;
262 memmove(s->buffer, s->buffer + len, s->buffer_index);
264 /* Get the next frame pts/duration */
265 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
274 #define OFFSET(x) offsetof(LAMEContext, x)
275 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
276 static const AVOption options[] = {
277 { "reservoir", "Use bit reservoir.", OFFSET(reservoir), AV_OPT_TYPE_INT, { 1 }, 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 .type = AVMEDIA_TYPE_AUDIO,
301 .priv_data_size = sizeof(LAMEContext),
302 .init = mp3lame_encode_init,
303 .encode2 = mp3lame_encode_frame,
304 .close = mp3lame_encode_close,
305 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME,
306 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32,
309 AV_SAMPLE_FMT_NONE },
310 .supported_samplerates = libmp3lame_sample_rates,
311 .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
312 .priv_class = &libmp3lame_class,
313 .defaults = libmp3lame_defaults,