blob: 5e26743f297dbb4dbb3afc0f1fb9f8d769ad34cb
1 | /* |
2 | * Interface to libmp3lame for mp3 encoding |
3 | * Copyright (c) 2002 Lennert Buytenhek <buytenh@gnu.org> |
4 | * |
5 | * This file is part of FFmpeg. |
6 | * |
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. |
11 | * |
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. |
16 | * |
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 |
20 | */ |
21 | |
22 | /** |
23 | * @file |
24 | * Interface to libmp3lame for mp3 encoding. |
25 | */ |
26 | |
27 | #include <lame/lame.h> |
28 | |
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" |
35 | #include "avcodec.h" |
36 | #include "audio_frame_queue.h" |
37 | #include "internal.h" |
38 | #include "mpegaudio.h" |
39 | #include "mpegaudiodecheader.h" |
40 | |
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. |
42 | |
43 | typedef struct LAMEContext { |
44 | AVClass *class; |
45 | AVCodecContext *avctx; |
46 | lame_global_flags *gfp; |
47 | uint8_t *buffer; |
48 | int buffer_index; |
49 | int buffer_size; |
50 | int reservoir; |
51 | int joint_stereo; |
52 | int abr; |
53 | int delay_sent; |
54 | float *samples_flt[2]; |
55 | AudioFrameQueue afq; |
56 | AVFloatDSPContext *fdsp; |
57 | } LAMEContext; |
58 | |
59 | |
60 | static int realloc_buffer(LAMEContext *s) |
61 | { |
62 | if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) { |
63 | int new_size = s->buffer_index + 2 * BUFFER_SIZE, err; |
64 | |
65 | ff_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size, |
66 | new_size); |
67 | if ((err = av_reallocp(&s->buffer, new_size)) < 0) { |
68 | s->buffer_size = s->buffer_index = 0; |
69 | return err; |
70 | } |
71 | s->buffer_size = new_size; |
72 | } |
73 | return 0; |
74 | } |
75 | |
76 | static av_cold int mp3lame_encode_close(AVCodecContext *avctx) |
77 | { |
78 | LAMEContext *s = avctx->priv_data; |
79 | |
80 | av_freep(&s->samples_flt[0]); |
81 | av_freep(&s->samples_flt[1]); |
82 | av_freep(&s->buffer); |
83 | av_freep(&s->fdsp); |
84 | |
85 | ff_af_queue_close(&s->afq); |
86 | |
87 | lame_close(s->gfp); |
88 | return 0; |
89 | } |
90 | |
91 | static av_cold int mp3lame_encode_init(AVCodecContext *avctx) |
92 | { |
93 | LAMEContext *s = avctx->priv_data; |
94 | int ret; |
95 | |
96 | s->avctx = avctx; |
97 | |
98 | /* initialize LAME and get defaults */ |
99 | if (!(s->gfp = lame_init())) |
100 | return AVERROR(ENOMEM); |
101 | |
102 | |
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); |
105 | |
106 | /* sample rate */ |
107 | lame_set_in_samplerate (s->gfp, avctx->sample_rate); |
108 | lame_set_out_samplerate(s->gfp, avctx->sample_rate); |
109 | |
110 | /* algorithmic quality */ |
111 | if (avctx->compression_level != FF_COMPRESSION_DEFAULT) |
112 | lame_set_quality(s->gfp, avctx->compression_level); |
113 | |
114 | /* rate control */ |
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); |
118 | } else { |
119 | if (avctx->bit_rate) { |
120 | if (s->abr) { // ABR |
121 | lame_set_VBR(s->gfp, vbr_abr); |
122 | lame_set_VBR_mean_bitrate_kbps(s->gfp, avctx->bit_rate / 1000); |
123 | } else // CBR |
124 | lame_set_brate(s->gfp, avctx->bit_rate / 1000); |
125 | } |
126 | } |
127 | |
128 | /* lowpass cutoff frequency */ |
129 | if (avctx->cutoff) |
130 | lame_set_lowpassfreq(s->gfp, avctx->cutoff); |
131 | |
132 | /* do not get a Xing VBR header frame from LAME */ |
133 | lame_set_bWriteVbrTag(s->gfp,0); |
134 | |
135 | /* bit reservoir usage */ |
136 | lame_set_disable_reservoir(s->gfp, !s->reservoir); |
137 | |
138 | /* set specified parameters */ |
139 | if (lame_init_params(s->gfp) < 0) { |
140 | ret = -1; |
141 | goto error; |
142 | } |
143 | |
144 | /* get encoder delay */ |
145 | avctx->initial_padding = lame_get_encoder_delay(s->gfp) + 528 + 1; |
146 | ff_af_queue_init(avctx, &s->afq); |
147 | |
148 | avctx->frame_size = lame_get_framesize(s->gfp); |
149 | |
150 | /* allocate float sample buffers */ |
151 | if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) { |
152 | int ch; |
153 | for (ch = 0; ch < avctx->channels; ch++) { |
154 | s->samples_flt[ch] = av_malloc_array(avctx->frame_size, |
155 | sizeof(*s->samples_flt[ch])); |
156 | if (!s->samples_flt[ch]) { |
157 | ret = AVERROR(ENOMEM); |
158 | goto error; |
159 | } |
160 | } |
161 | } |
162 | |
163 | ret = realloc_buffer(s); |
164 | if (ret < 0) |
165 | goto error; |
166 | |
167 | s->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT); |
168 | if (!s->fdsp) { |
169 | ret = AVERROR(ENOMEM); |
170 | goto error; |
171 | } |
172 | |
173 | |
174 | return 0; |
175 | error: |
176 | mp3lame_encode_close(avctx); |
177 | return ret; |
178 | } |
179 | |
180 | #define ENCODE_BUFFER(func, buf_type, buf_name) do { \ |
181 | lame_result = func(s->gfp, \ |
182 | (const buf_type *)buf_name[0], \ |
183 | (const buf_type *)buf_name[1], frame->nb_samples, \ |
184 | s->buffer + s->buffer_index, \ |
185 | s->buffer_size - s->buffer_index); \ |
186 | } while (0) |
187 | |
188 | static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, |
189 | const AVFrame *frame, int *got_packet_ptr) |
190 | { |
191 | LAMEContext *s = avctx->priv_data; |
192 | MPADecodeHeader hdr; |
193 | int len, ret, ch, discard_padding; |
194 | int lame_result; |
195 | uint32_t h; |
196 | |
197 | if (frame) { |
198 | switch (avctx->sample_fmt) { |
199 | case AV_SAMPLE_FMT_S16P: |
200 | ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data); |
201 | break; |
202 | case AV_SAMPLE_FMT_S32P: |
203 | ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data); |
204 | break; |
205 | case AV_SAMPLE_FMT_FLTP: |
206 | if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) { |
207 | av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n"); |
208 | return AVERROR(EINVAL); |
209 | } |
210 | for (ch = 0; ch < avctx->channels; ch++) { |
211 | s->fdsp->vector_fmul_scalar(s->samples_flt[ch], |
212 | (const float *)frame->data[ch], |
213 | 32768.0f, |
214 | FFALIGN(frame->nb_samples, 8)); |
215 | } |
216 | ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt); |
217 | break; |
218 | default: |
219 | return AVERROR_BUG; |
220 | } |
221 | } else if (!s->afq.frame_alloc) { |
222 | lame_result = 0; |
223 | } else { |
224 | lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index, |
225 | s->buffer_size - s->buffer_index); |
226 | } |
227 | if (lame_result < 0) { |
228 | if (lame_result == -1) { |
229 | av_log(avctx, AV_LOG_ERROR, |
230 | "lame: output buffer too small (buffer index: %d, free bytes: %d)\n", |
231 | s->buffer_index, s->buffer_size - s->buffer_index); |
232 | } |
233 | return -1; |
234 | } |
235 | s->buffer_index += lame_result; |
236 | ret = realloc_buffer(s); |
237 | if (ret < 0) { |
238 | av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n"); |
239 | return ret; |
240 | } |
241 | |
242 | /* add current frame to the queue */ |
243 | if (frame) { |
244 | if ((ret = ff_af_queue_add(&s->afq, frame)) < 0) |
245 | return ret; |
246 | } |
247 | |
248 | /* Move 1 frame from the LAME buffer to the output packet, if available. |
249 | We have to parse the first frame header in the output buffer to |
250 | determine the frame size. */ |
251 | if (s->buffer_index < 4) |
252 | return 0; |
253 | h = AV_RB32(s->buffer); |
254 | |
255 | ret = avpriv_mpegaudio_decode_header(&hdr, h); |
256 | if (ret < 0) { |
257 | av_log(avctx, AV_LOG_ERROR, "Invalid mp3 header at start of buffer\n"); |
258 | return AVERROR_BUG; |
259 | } else if (ret) { |
260 | av_log(avctx, AV_LOG_ERROR, "free format output not supported\n"); |
261 | return -1; |
262 | } |
263 | len = hdr.frame_size; |
264 | ff_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len, |
265 | s->buffer_index); |
266 | if (len <= s->buffer_index) { |
267 | if ((ret = ff_alloc_packet2(avctx, avpkt, len, 0)) < 0) |
268 | return ret; |
269 | memcpy(avpkt->data, s->buffer, len); |
270 | s->buffer_index -= len; |
271 | memmove(s->buffer, s->buffer + len, s->buffer_index); |
272 | |
273 | /* Get the next frame pts/duration */ |
274 | ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts, |
275 | &avpkt->duration); |
276 | |
277 | discard_padding = avctx->frame_size - avpkt->duration; |
278 | // Check if subtraction resulted in an overflow |
279 | if ((discard_padding < avctx->frame_size) != (avpkt->duration > 0)) { |
280 | av_log(avctx, AV_LOG_ERROR, "discard padding overflow\n"); |
281 | av_packet_unref(avpkt); |
282 | av_free(avpkt); |
283 | return AVERROR(EINVAL); |
284 | } |
285 | if ((!s->delay_sent && avctx->initial_padding > 0) || discard_padding > 0) { |
286 | uint8_t* side_data = av_packet_new_side_data(avpkt, |
287 | AV_PKT_DATA_SKIP_SAMPLES, |
288 | 10); |
289 | if(!side_data) { |
290 | av_packet_unref(avpkt); |
291 | av_free(avpkt); |
292 | return AVERROR(ENOMEM); |
293 | } |
294 | if (!s->delay_sent) { |
295 | AV_WL32(side_data, avctx->initial_padding); |
296 | s->delay_sent = 1; |
297 | } |
298 | AV_WL32(side_data + 4, discard_padding); |
299 | } |
300 | |
301 | avpkt->size = len; |
302 | *got_packet_ptr = 1; |
303 | } |
304 | return 0; |
305 | } |
306 | |
307 | #define OFFSET(x) offsetof(LAMEContext, x) |
308 | #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM |
309 | static const AVOption options[] = { |
310 | { "reservoir", "use bit reservoir", OFFSET(reservoir), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, AE }, |
311 | { "joint_stereo", "use joint stereo", OFFSET(joint_stereo), AV_OPT_TYPE_BOOL, { .i64 = 1 }, 0, 1, AE }, |
312 | { "abr", "use ABR", OFFSET(abr), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, AE }, |
313 | { NULL }, |
314 | }; |
315 | |
316 | static const AVClass libmp3lame_class = { |
317 | .class_name = "libmp3lame encoder", |
318 | .item_name = av_default_item_name, |
319 | .option = options, |
320 | .version = LIBAVUTIL_VERSION_INT, |
321 | }; |
322 | |
323 | static const AVCodecDefault libmp3lame_defaults[] = { |
324 | { "b", "0" }, |
325 | { NULL }, |
326 | }; |
327 | |
328 | static const int libmp3lame_sample_rates[] = { |
329 | 44100, 48000, 32000, 22050, 24000, 16000, 11025, 12000, 8000, 0 |
330 | }; |
331 | |
332 | AVCodec ff_libmp3lame_encoder = { |
333 | .name = "libmp3lame", |
334 | .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"), |
335 | .type = AVMEDIA_TYPE_AUDIO, |
336 | .id = AV_CODEC_ID_MP3, |
337 | .priv_data_size = sizeof(LAMEContext), |
338 | .init = mp3lame_encode_init, |
339 | .encode2 = mp3lame_encode_frame, |
340 | .close = mp3lame_encode_close, |
341 | .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_SMALL_LAST_FRAME, |
342 | .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P, |
343 | AV_SAMPLE_FMT_FLTP, |
344 | AV_SAMPLE_FMT_S16P, |
345 | AV_SAMPLE_FMT_NONE }, |
346 | .supported_samplerates = libmp3lame_sample_rates, |
347 | .channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO, |
348 | AV_CH_LAYOUT_STEREO, |
349 | 0 }, |
350 | .priv_class = &libmp3lame_class, |
351 | .defaults = libmp3lame_defaults, |
352 | }; |
353 |