Commit | Line | Data |
---|---|---|
04d7f601 DB |
1 | /* |
2 | * copyright (c) 2002 Mark Hills <mark@pogo.org.uk> | |
3 | * | |
2912e87a | 4 | * This file is part of Libav. |
b78e7197 | 5 | * |
2912e87a | 6 | * Libav is free software; you can redistribute it and/or |
04d7f601 DB |
7 | * modify it under the terms of the GNU Lesser General Public |
8 | * License as published by the Free Software Foundation; either | |
b78e7197 | 9 | * version 2.1 of the License, or (at your option) any later version. |
04d7f601 | 10 | * |
2912e87a | 11 | * Libav is distributed in the hope that it will be useful, |
04d7f601 DB |
12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
14 | * Lesser General Public License for more details. | |
15 | * | |
16 | * You should have received a copy of the GNU Lesser General Public | |
2912e87a | 17 | * License along with Libav; if not, write to the Free Software |
e5a389a1 | 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
04d7f601 DB |
19 | */ |
20 | ||
983e3246 | 21 | /** |
ba87f080 | 22 | * @file |
983e3246 MN |
23 | * Ogg Vorbis codec support via libvorbisenc. |
24 | * @author Mark Hills <mark@pogo.org.uk> | |
81e0d0b4 MH |
25 | */ |
26 | ||
81e0d0b4 MH |
27 | #include <vorbis/vorbisenc.h> |
28 | ||
77336a5e | 29 | #include "libavutil/opt.h" |
81e0d0b4 | 30 | #include "avcodec.h" |
2c124cb6 | 31 | #include "bytestream.h" |
9577838f | 32 | #include "vorbis.h" |
954a6532 | 33 | #include "libavutil/mathematics.h" |
81e0d0b4 | 34 | |
3f4993f1 MN |
35 | #undef NDEBUG |
36 | #include <assert.h> | |
37 | ||
6d8f985e | 38 | #define OGGVORBIS_FRAME_SIZE 64 |
81e0d0b4 | 39 | |
ca5ab8cd | 40 | #define BUFFER_SIZE (1024 * 64) |
81e0d0b4 MH |
41 | |
42 | typedef struct OggVorbisContext { | |
77336a5e | 43 | AVClass *av_class; |
ca5ab8cd DB |
44 | vorbis_info vi; |
45 | vorbis_dsp_state vd; | |
46 | vorbis_block vb; | |
6d8f985e MN |
47 | uint8_t buffer[BUFFER_SIZE]; |
48 | int buffer_index; | |
9c8f0768 | 49 | int eof; |
3aca208a MH |
50 | |
51 | /* decoder */ | |
ca5ab8cd | 52 | vorbis_comment vc; |
d76f581f | 53 | ogg_packet op; |
77336a5e MN |
54 | |
55 | double iblock; | |
ca5ab8cd | 56 | } OggVorbisContext; |
81e0d0b4 | 57 | |
ca5ab8cd DB |
58 | static const AVOption options[] = { |
59 | { "iblock", "Sets the impulse block bias", offsetof(OggVorbisContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM }, | |
60 | { NULL } | |
77336a5e | 61 | }; |
da754858 | 62 | static const AVClass class = { "libvorbis", av_default_item_name, options, LIBAVUTIL_VERSION_INT }; |
81e0d0b4 | 63 | |
ca5ab8cd DB |
64 | static av_cold int oggvorbis_init_encoder(vorbis_info *vi, AVCodecContext *avccontext) |
65 | { | |
66 | OggVorbisContext *context = avccontext->priv_data; | |
13c71451 | 67 | double cfreq; |
c55427f8 | 68 | |
ca5ab8cd | 69 | if (avccontext->flags & CODEC_FLAG_QSCALE) { |
13c71451 | 70 | /* variable bitrate */ |
ca5ab8cd DB |
71 | if (vorbis_encode_setup_vbr(vi, avccontext->channels, |
72 | avccontext->sample_rate, | |
73 | avccontext->global_quality / (float)FF_QP2LAMBDA / 10.0)) | |
13c71451 JR |
74 | return -1; |
75 | } else { | |
e5a5ea9e DC |
76 | int minrate = avccontext->rc_min_rate > 0 ? avccontext->rc_min_rate : -1; |
77 | int maxrate = avccontext->rc_min_rate > 0 ? avccontext->rc_max_rate : -1; | |
78 | ||
13c71451 | 79 | /* constant bitrate */ |
ca5ab8cd DB |
80 | if (vorbis_encode_setup_managed(vi, avccontext->channels, |
81 | avccontext->sample_rate, minrate, | |
82 | avccontext->bit_rate, maxrate)) | |
13c71451 JR |
83 | return -1; |
84 | ||
57ebbccf | 85 | /* variable bitrate by estimate, disable slow rate management */ |
ca5ab8cd DB |
86 | if (minrate == -1 && maxrate == -1) |
87 | if (vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)) | |
57ebbccf | 88 | return -1; |
13c71451 | 89 | } |
c55427f8 | 90 | |
13c71451 | 91 | /* cutoff frequency */ |
ca5ab8cd | 92 | if (avccontext->cutoff > 0) { |
13c71451 | 93 | cfreq = avccontext->cutoff / 1000.0; |
ca5ab8cd | 94 | if (vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)) |
13c71451 JR |
95 | return -1; |
96 | } | |
81e0d0b4 | 97 | |
ca5ab8cd | 98 | if (context->iblock) { |
77336a5e MN |
99 | vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &context->iblock); |
100 | } | |
101 | ||
13c71451 | 102 | return vorbis_encode_setup_init(vi); |
81e0d0b4 MH |
103 | } |
104 | ||
fd7242dd | 105 | /* How many bytes are needed for a buffer of length 'l' */ |
ca5ab8cd DB |
106 | static int xiph_len(int l) |
107 | { | |
108 | return (1 + l / 255 + l); | |
109 | } | |
fd7242dd | 110 | |
ca5ab8cd DB |
111 | static av_cold int oggvorbis_encode_init(AVCodecContext *avccontext) |
112 | { | |
113 | OggVorbisContext *context = avccontext->priv_data; | |
bbb77e7c MN |
114 | ogg_packet header, header_comm, header_code; |
115 | uint8_t *p; | |
fd7242dd | 116 | unsigned int offset; |
81e0d0b4 | 117 | |
ca5ab8cd DB |
118 | vorbis_info_init(&context->vi); |
119 | if (oggvorbis_init_encoder(&context->vi, avccontext) < 0) { | |
120 | av_log(avccontext, AV_LOG_ERROR, "oggvorbis_encode_init: init_encoder failed\n"); | |
121 | return -1; | |
81e0d0b4 | 122 | } |
ca5ab8cd DB |
123 | vorbis_analysis_init(&context->vd, &context->vi); |
124 | vorbis_block_init(&context->vd, &context->vb); | |
81e0d0b4 | 125 | |
bbb77e7c | 126 | vorbis_comment_init(&context->vc); |
ca5ab8cd | 127 | vorbis_comment_add_tag(&context->vc, "encoder", LIBAVCODEC_IDENT); |
bbb77e7c MN |
128 | |
129 | vorbis_analysis_headerout(&context->vd, &context->vc, &header, | |
ca5ab8cd | 130 | &header_comm, &header_code); |
115329f1 | 131 | |
ca5ab8cd | 132 | avccontext->extradata_size = |
fd7242dd PM |
133 | 1 + xiph_len(header.bytes) + xiph_len(header_comm.bytes) + |
134 | header_code.bytes; | |
135 | p = avccontext->extradata = | |
ca5ab8cd DB |
136 | av_malloc(avccontext->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE); |
137 | p[0] = 2; | |
138 | offset = 1; | |
ad2b531d MR |
139 | offset += av_xiphlacing(&p[offset], header.bytes); |
140 | offset += av_xiphlacing(&p[offset], header_comm.bytes); | |
141 | memcpy(&p[offset], header.packet, header.bytes); | |
142 | offset += header.bytes; | |
143 | memcpy(&p[offset], header_comm.packet, header_comm.bytes); | |
144 | offset += header_comm.bytes; | |
145 | memcpy(&p[offset], header_code.packet, header_code.bytes); | |
146 | offset += header_code.bytes; | |
fd7242dd | 147 | assert(offset == avccontext->extradata_size); |
115329f1 | 148 | |
ca5ab8cd DB |
149 | #if 0 |
150 | vorbis_block_clear(&context->vb); | |
bbb77e7c | 151 | vorbis_dsp_clear(&context->vd); |
ca5ab8cd DB |
152 | vorbis_info_clear(&context->vi); |
153 | #endif | |
bbb77e7c | 154 | vorbis_comment_clear(&context->vc); |
115329f1 | 155 | |
ca5ab8cd | 156 | avccontext->frame_size = OGGVORBIS_FRAME_SIZE; |
115329f1 | 157 | |
ca5ab8cd DB |
158 | avccontext->coded_frame = avcodec_alloc_frame(); |
159 | avccontext->coded_frame->key_frame = 1; | |
115329f1 | 160 | |
ca5ab8cd | 161 | return 0; |
81e0d0b4 MH |
162 | } |
163 | ||
3aca208a | 164 | static int oggvorbis_encode_frame(AVCodecContext *avccontext, |
bb270c08 | 165 | unsigned char *packets, |
ca5ab8cd | 166 | int buf_size, void *data) |
81e0d0b4 | 167 | { |
ca5ab8cd DB |
168 | OggVorbisContext *context = avccontext->priv_data; |
169 | ogg_packet op; | |
170 | signed short *audio = data; | |
9c8f0768 NG |
171 | int l; |
172 | ||
ca5ab8cd | 173 | if (data) { |
42859ddb | 174 | const int samples = avccontext->frame_size; |
ca5ab8cd | 175 | float **buffer; |
9577838f | 176 | int c, channels = context->vi.channels; |
81e0d0b4 | 177 | |
ca5ab8cd | 178 | buffer = vorbis_analysis_buffer(&context->vd, samples); |
9577838f JD |
179 | for (c = 0; c < channels; c++) { |
180 | int co = (channels > 8) ? c : | |
ca5ab8cd DB |
181 | ff_vorbis_encoding_channel_layout_offsets[channels - 1][c]; |
182 | for (l = 0; l < samples; l++) | |
183 | buffer[c][l] = audio[l * channels + co] / 32768.f; | |
bb270c08 | 184 | } |
ca5ab8cd | 185 | vorbis_analysis_wrote(&context->vd, samples); |
9c8f0768 | 186 | } else { |
ca5ab8cd DB |
187 | if (!context->eof) |
188 | vorbis_analysis_wrote(&context->vd, 0); | |
9c8f0768 NG |
189 | context->eof = 1; |
190 | } | |
81e0d0b4 | 191 | |
ca5ab8cd | 192 | while (vorbis_analysis_blockout(&context->vd, &context->vb) == 1) { |
bb270c08 | 193 | vorbis_analysis(&context->vb, NULL); |
ca5ab8cd | 194 | vorbis_bitrate_addblock(&context->vb); |
81e0d0b4 | 195 | |
ca5ab8cd | 196 | while (vorbis_bitrate_flushpacket(&context->vd, &op)) { |
0afd2a92 DB |
197 | /* i'd love to say the following line is a hack, but sadly it's |
198 | * not, apparently the end of stream decision is in libogg. */ | |
ca5ab8cd | 199 | if (op.bytes == 1 && op.e_o_s) |
6f824977 | 200 | continue; |
c426562c PM |
201 | if (context->buffer_index + sizeof(ogg_packet) + op.bytes > BUFFER_SIZE) { |
202 | av_log(avccontext, AV_LOG_ERROR, "libvorbis: buffer overflow."); | |
203 | return -1; | |
204 | } | |
6d8f985e MN |
205 | memcpy(context->buffer + context->buffer_index, &op, sizeof(ogg_packet)); |
206 | context->buffer_index += sizeof(ogg_packet); | |
207 | memcpy(context->buffer + context->buffer_index, op.packet, op.bytes); | |
208 | context->buffer_index += op.bytes; | |
209 | // av_log(avccontext, AV_LOG_DEBUG, "e%d / %d\n", context->buffer_index, op.bytes); | |
bb270c08 | 210 | } |
81e0d0b4 MH |
211 | } |
212 | ||
ca5ab8cd DB |
213 | l = 0; |
214 | if (context->buffer_index) { | |
215 | ogg_packet *op2 = (ogg_packet *)context->buffer; | |
6d8f985e | 216 | op2->packet = context->buffer + sizeof(ogg_packet); |
3f4993f1 | 217 | |
ca5ab8cd DB |
218 | l = op2->bytes; |
219 | avccontext->coded_frame->pts = av_rescale_q(op2->granulepos, (AVRational) { 1, avccontext->sample_rate }, avccontext->time_base); | |
3723f01c | 220 | //FIXME we should reorder the user supplied pts and not assume that they are spaced by 1/sample_rate |
3f4993f1 | 221 | |
c426562c PM |
222 | if (l > buf_size) { |
223 | av_log(avccontext, AV_LOG_ERROR, "libvorbis: buffer overflow."); | |
224 | return -1; | |
225 | } | |
226 | ||
6f824977 MN |
227 | memcpy(packets, op2->packet, l); |
228 | context->buffer_index -= l + sizeof(ogg_packet); | |
1204a13c | 229 | memmove(context->buffer, context->buffer + l + sizeof(ogg_packet), context->buffer_index); |
6d8f985e | 230 | // av_log(avccontext, AV_LOG_DEBUG, "E%d\n", l); |
6d8f985e MN |
231 | } |
232 | ||
3f4993f1 | 233 | return l; |
81e0d0b4 MH |
234 | } |
235 | ||
ca5ab8cd DB |
236 | static av_cold int oggvorbis_encode_close(AVCodecContext *avccontext) |
237 | { | |
238 | OggVorbisContext *context = avccontext->priv_data; | |
81e0d0b4 | 239 | /* ogg_packet op ; */ |
115329f1 | 240 | |
ca5ab8cd | 241 | vorbis_analysis_wrote(&context->vd, 0); /* notify vorbisenc this is EOF */ |
81e0d0b4 | 242 | |
81e0d0b4 MH |
243 | vorbis_block_clear(&context->vb); |
244 | vorbis_dsp_clear(&context->vd); | |
245 | vorbis_info_clear(&context->vi); | |
492cd3a9 MN |
246 | |
247 | av_freep(&avccontext->coded_frame); | |
bbb77e7c | 248 | av_freep(&avccontext->extradata); |
115329f1 | 249 | |
ca5ab8cd | 250 | return 0; |
81e0d0b4 MH |
251 | } |
252 | ||
d36beb3f | 253 | AVCodec ff_libvorbis_encoder = { |
86714887 DB |
254 | .name = "libvorbis", |
255 | .type = AVMEDIA_TYPE_AUDIO, | |
256 | .id = CODEC_ID_VORBIS, | |
257 | .priv_data_size = sizeof(OggVorbisContext), | |
258 | .init = oggvorbis_encode_init, | |
259 | .encode = oggvorbis_encode_frame, | |
260 | .close = oggvorbis_encode_close, | |
261 | .capabilities = CODEC_CAP_DELAY, | |
ca5ab8cd | 262 | .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE }, |
86714887 DB |
263 | .long_name = NULL_IF_CONFIG_SMALL("libvorbis Vorbis"), |
264 | .priv_class = &class, | |
265 | }; |