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;
54 float *samples_flt[2];
56 AVFloatDSPContext *fdsp;
60 static int realloc_buffer(LAMEContext *s)
62 if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) {
63 int new_size = s->buffer_index + 2 * BUFFER_SIZE, err;
65 ff_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size,
67 if ((err = av_reallocp(&s->buffer, new_size)) < 0) {
68 s->buffer_size = s->buffer_index = 0;
71 s->buffer_size = new_size;
76 static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
78 LAMEContext *s = avctx->priv_data;
80 av_freep(&s->samples_flt[0]);
81 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()))
100 return AVERROR(ENOMEM);
103 lame_set_num_channels(s->gfp, avctx->channels);
104 lame_set_mode(s->gfp, avctx->channels > 1 ? s->joint_stereo ? JOINT_STEREO : STEREO : MONO);
107 lame_set_in_samplerate (s->gfp, avctx->sample_rate);
108 lame_set_out_samplerate(s->gfp, avctx->sample_rate);
110 /* algorithmic quality */
111 if (avctx->compression_level != FF_COMPRESSION_DEFAULT)
112 lame_set_quality(s->gfp, avctx->compression_level);
115 if (avctx->flags & AV_CODEC_FLAG_QSCALE) { // VBR
116 lame_set_VBR(s->gfp, vbr_default);
117 lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
119 if (avctx->bit_rate) {
121 lame_set_VBR(s->gfp, vbr_abr);
122 lame_set_VBR_mean_bitrate_kbps(s->gfp, avctx->bit_rate / 1000);
124 lame_set_brate(s->gfp, avctx->bit_rate / 1000);
128 /* do not get a Xing VBR header frame from LAME */
129 lame_set_bWriteVbrTag(s->gfp,0);
131 /* bit reservoir usage */
132 lame_set_disable_reservoir(s->gfp, !s->reservoir);
134 /* set specified parameters */
135 if (lame_init_params(s->gfp) < 0) {
140 /* get encoder delay */
141 avctx->initial_padding = lame_get_encoder_delay(s->gfp) + 528 + 1;
142 ff_af_queue_init(avctx, &s->afq);
144 avctx->frame_size = lame_get_framesize(s->gfp);
146 /* allocate float sample buffers */
147 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
149 for (ch = 0; ch < avctx->channels; ch++) {
150 s->samples_flt[ch] = av_malloc_array(avctx->frame_size,
151 sizeof(*s->samples_flt[ch]));
152 if (!s->samples_flt[ch]) {
153 ret = AVERROR(ENOMEM);
159 ret = realloc_buffer(s);
163 s->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
165 ret = AVERROR(ENOMEM);
172 mp3lame_encode_close(avctx);
176 #define ENCODE_BUFFER(func, buf_type, buf_name) do { \
177 lame_result = func(s->gfp, \
178 (const buf_type *)buf_name[0], \
179 (const buf_type *)buf_name[1], frame->nb_samples, \
180 s->buffer + s->buffer_index, \
181 s->buffer_size - s->buffer_index); \
184 static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
185 const AVFrame *frame, int *got_packet_ptr)
187 LAMEContext *s = avctx->priv_data;
189 int len, ret, ch, discard_padding;
194 switch (avctx->sample_fmt) {
195 case AV_SAMPLE_FMT_S16P:
196 ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data);
198 case AV_SAMPLE_FMT_S32P:
199 ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data);
201 case AV_SAMPLE_FMT_FLTP:
202 if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) {
203 av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n");
204 return AVERROR(EINVAL);
206 for (ch = 0; ch < avctx->channels; ch++) {
207 s->fdsp->vector_fmul_scalar(s->samples_flt[ch],
208 (const float *)frame->data[ch],
210 FFALIGN(frame->nb_samples, 8));
212 ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt);
217 } else if (!s->afq.frame_alloc) {
220 lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
221 s->buffer_size - s->buffer_index);
223 if (lame_result < 0) {
224 if (lame_result == -1) {
225 av_log(avctx, AV_LOG_ERROR,
226 "lame: output buffer too small (buffer index: %d, free bytes: %d)\n",
227 s->buffer_index, s->buffer_size - s->buffer_index);
231 s->buffer_index += lame_result;
232 ret = realloc_buffer(s);
234 av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n");
238 /* add current frame to the queue */
240 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
244 /* Move 1 frame from the LAME buffer to the output packet, if available.
245 We have to parse the first frame header in the output buffer to
246 determine the frame size. */
247 if (s->buffer_index < 4)
249 h = AV_RB32(s->buffer);
251 ret = avpriv_mpegaudio_decode_header(&hdr, h);
253 av_log(avctx, AV_LOG_ERROR, "Invalid mp3 header at start of buffer\n");
256 av_log(avctx, AV_LOG_ERROR, "free format output not supported\n");
259 len = hdr.frame_size;
260 ff_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
262 if (len <= s->buffer_index) {
263 if ((ret = ff_alloc_packet2(avctx, avpkt, len, 0)) < 0)
265 memcpy(avpkt->data, s->buffer, len);
266 s->buffer_index -= len;
267 memmove(s->buffer, s->buffer + len, s->buffer_index);
269 /* Get the next frame pts/duration */
270 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
273 discard_padding = avctx->frame_size - avpkt->duration;
274 // Check if subtraction resulted in an overflow
275 if ((discard_padding < avctx->frame_size) != (avpkt->duration > 0)) {
276 av_log(avctx, AV_LOG_ERROR, "discard padding overflow\n");
277 av_packet_unref(avpkt);
279 return AVERROR(EINVAL);
281 if ((!s->delay_sent && avctx->initial_padding > 0) || discard_padding > 0) {
282 uint8_t* side_data = av_packet_new_side_data(avpkt,
283 AV_PKT_DATA_SKIP_SAMPLES,
286 av_packet_unref(avpkt);
288 return AVERROR(ENOMEM);
290 if (!s->delay_sent) {
291 AV_WL32(side_data, avctx->initial_padding);
294 AV_WL32(side_data + 4, discard_padding);
303 #define OFFSET(x) offsetof(LAMEContext, x)
304 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
305 static const AVOption options[] = {
306 { "reservoir", "use bit reservoir", OFFSET(reservoir), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, AE },
307 { "joint_stereo", "use joint stereo", OFFSET(joint_stereo), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, AE },
308 { "abr", "use ABR", OFFSET(abr), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, AE },
312 static const AVClass libmp3lame_class = {
313 .class_name = "libmp3lame encoder",
314 .item_name = av_default_item_name,
316 .version = LIBAVUTIL_VERSION_INT,
319 static const AVCodecDefault libmp3lame_defaults[] = {
324 static const int libmp3lame_sample_rates[] = {
325 44100, 48000, 32000, 22050, 24000, 16000, 11025, 12000, 8000, 0
328 AVCodec ff_libmp3lame_encoder = {
329 .name = "libmp3lame",
330 .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
331 .type = AVMEDIA_TYPE_AUDIO,
332 .id = AV_CODEC_ID_MP3,
333 .priv_data_size = sizeof(LAMEContext),
334 .init = mp3lame_encode_init,
335 .encode2 = mp3lame_encode_frame,
336 .close = mp3lame_encode_close,
337 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_SMALL_LAST_FRAME,
338 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P,
341 AV_SAMPLE_FMT_NONE },
342 .supported_samplerates = libmp3lame_sample_rates,
343 .channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
346 .priv_class = &libmp3lame_class,
347 .defaults = libmp3lame_defaults,