Commit | Line | Data |
---|---|---|
983e3246 MN |
1 | /** |
2 | * @file oggvorbis.c | |
3 | * Ogg Vorbis codec support via libvorbisenc. | |
4 | * @author Mark Hills <mark@pogo.org.uk> | |
81e0d0b4 MH |
5 | */ |
6 | ||
81e0d0b4 MH |
7 | #include <vorbis/vorbisenc.h> |
8 | ||
9 | #include "avcodec.h" | |
81e0d0b4 | 10 | |
3f4993f1 MN |
11 | #undef NDEBUG |
12 | #include <assert.h> | |
13 | ||
6d8f985e | 14 | #define OGGVORBIS_FRAME_SIZE 64 |
81e0d0b4 | 15 | |
6d8f985e | 16 | #define BUFFER_SIZE (1024*64) |
81e0d0b4 MH |
17 | |
18 | typedef struct OggVorbisContext { | |
19 | vorbis_info vi ; | |
20 | vorbis_dsp_state vd ; | |
21 | vorbis_block vb ; | |
6d8f985e MN |
22 | uint8_t buffer[BUFFER_SIZE]; |
23 | int buffer_index; | |
3aca208a MH |
24 | |
25 | /* decoder */ | |
26 | vorbis_comment vc ; | |
d76f581f | 27 | ogg_packet op; |
81e0d0b4 MH |
28 | } OggVorbisContext ; |
29 | ||
30 | ||
bbb77e7c | 31 | static int oggvorbis_init_encoder(vorbis_info *vi, AVCodecContext *avccontext) { |
c55427f8 | 32 | |
1ff8f2b3 JR |
33 | if(avccontext->flags & CODEC_FLAG_QSCALE) { |
34 | return vorbis_encode_init_vbr(vi, avccontext->channels, | |
35 | avccontext->sample_rate, | |
36 | avccontext->global_quality / (float)FF_QP2LAMBDA); | |
37 | } | |
c55427f8 MH |
38 | #ifdef OGGVORBIS_VBR_BY_ESTIMATE |
39 | /* variable bitrate by estimate */ | |
40 | ||
41 | return (vorbis_encode_setup_managed(vi, avccontext->channels, | |
42 | avccontext->sample_rate, -1, avccontext->bit_rate, -1) || | |
43 | vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE_AVG, NULL) || | |
44 | vorbis_encode_setup_init(vi)) ; | |
45 | #else | |
46 | /* constant bitrate */ | |
81e0d0b4 MH |
47 | |
48 | return vorbis_encode_init(vi, avccontext->channels, | |
49 | avccontext->sample_rate, -1, avccontext->bit_rate, -1) ; | |
c55427f8 | 50 | #endif |
81e0d0b4 MH |
51 | } |
52 | ||
81e0d0b4 MH |
53 | static int oggvorbis_encode_init(AVCodecContext *avccontext) { |
54 | OggVorbisContext *context = avccontext->priv_data ; | |
bbb77e7c MN |
55 | ogg_packet header, header_comm, header_code; |
56 | uint8_t *p; | |
ad2b531d | 57 | unsigned int offset, len; |
81e0d0b4 | 58 | |
81e0d0b4 | 59 | vorbis_info_init(&context->vi) ; |
81e0d0b4 | 60 | if(oggvorbis_init_encoder(&context->vi, avccontext) < 0) { |
9b879566 | 61 | av_log(avccontext, AV_LOG_ERROR, "oggvorbis_encode_init: init_encoder failed") ; |
81e0d0b4 MH |
62 | return -1 ; |
63 | } | |
81e0d0b4 MH |
64 | vorbis_analysis_init(&context->vd, &context->vi) ; |
65 | vorbis_block_init(&context->vd, &context->vb) ; | |
66 | ||
bbb77e7c MN |
67 | vorbis_comment_init(&context->vc); |
68 | vorbis_comment_add_tag(&context->vc, "encoder", LIBAVCODEC_IDENT) ; | |
69 | ||
70 | vorbis_analysis_headerout(&context->vd, &context->vc, &header, | |
71 | &header_comm, &header_code); | |
115329f1 | 72 | |
ad2b531d MR |
73 | len = header.bytes + header_comm.bytes + header_code.bytes; |
74 | avccontext->extradata_size= 64 + len + len/255; | |
75 | p = avccontext->extradata= av_mallocz(avccontext->extradata_size); | |
76 | p[0] = 2; | |
77 | offset = 1; | |
78 | offset += av_xiphlacing(&p[offset], header.bytes); | |
79 | offset += av_xiphlacing(&p[offset], header_comm.bytes); | |
80 | memcpy(&p[offset], header.packet, header.bytes); | |
81 | offset += header.bytes; | |
82 | memcpy(&p[offset], header_comm.packet, header_comm.bytes); | |
83 | offset += header_comm.bytes; | |
84 | memcpy(&p[offset], header_code.packet, header_code.bytes); | |
85 | offset += header_code.bytes; | |
86 | avccontext->extradata_size = offset; | |
87 | avccontext->extradata= av_realloc(avccontext->extradata, avccontext->extradata_size); | |
115329f1 | 88 | |
bbb77e7c MN |
89 | /* vorbis_block_clear(&context->vb); |
90 | vorbis_dsp_clear(&context->vd); | |
91 | vorbis_info_clear(&context->vi);*/ | |
92 | vorbis_comment_clear(&context->vc); | |
115329f1 | 93 | |
81e0d0b4 | 94 | avccontext->frame_size = OGGVORBIS_FRAME_SIZE ; |
115329f1 | 95 | |
492cd3a9 MN |
96 | avccontext->coded_frame= avcodec_alloc_frame(); |
97 | avccontext->coded_frame->key_frame= 1; | |
115329f1 | 98 | |
81e0d0b4 MH |
99 | return 0 ; |
100 | } | |
101 | ||
102 | ||
3aca208a MH |
103 | static int oggvorbis_encode_frame(AVCodecContext *avccontext, |
104 | unsigned char *packets, | |
81e0d0b4 MH |
105 | int buf_size, void *data) |
106 | { | |
107 | OggVorbisContext *context = avccontext->priv_data ; | |
108 | float **buffer ; | |
109 | ogg_packet op ; | |
0d00cf38 | 110 | signed short *audio = data ; |
6f824977 | 111 | int l, samples = data ? OGGVORBIS_FRAME_SIZE : 0; |
81e0d0b4 MH |
112 | |
113 | buffer = vorbis_analysis_buffer(&context->vd, samples) ; | |
114 | ||
115 | if(context->vi.channels == 1) { | |
116 | for(l = 0 ; l < samples ; l++) | |
0d00cf38 | 117 | buffer[0][l]=audio[l]/32768.f; |
81e0d0b4 MH |
118 | } else { |
119 | for(l = 0 ; l < samples ; l++){ | |
0d00cf38 SS |
120 | buffer[0][l]=audio[l*2]/32768.f; |
121 | buffer[1][l]=audio[l*2+1]/32768.f; | |
81e0d0b4 MH |
122 | } |
123 | } | |
115329f1 DB |
124 | |
125 | vorbis_analysis_wrote(&context->vd, samples) ; | |
81e0d0b4 | 126 | |
81e0d0b4 MH |
127 | while(vorbis_analysis_blockout(&context->vd, &context->vb) == 1) { |
128 | vorbis_analysis(&context->vb, NULL); | |
129 | vorbis_bitrate_addblock(&context->vb) ; | |
130 | ||
131 | while(vorbis_bitrate_flushpacket(&context->vd, &op)) { | |
6f824977 MN |
132 | if(op.bytes==1) //id love to say this is a hack, bad sadly its not, appearently the end of stream decission is in libogg |
133 | continue; | |
6d8f985e MN |
134 | memcpy(context->buffer + context->buffer_index, &op, sizeof(ogg_packet)); |
135 | context->buffer_index += sizeof(ogg_packet); | |
136 | memcpy(context->buffer + context->buffer_index, op.packet, op.bytes); | |
137 | context->buffer_index += op.bytes; | |
138 | // av_log(avccontext, AV_LOG_DEBUG, "e%d / %d\n", context->buffer_index, op.bytes); | |
81e0d0b4 MH |
139 | } |
140 | } | |
141 | ||
3f4993f1 | 142 | l=0; |
6d8f985e | 143 | if(context->buffer_index){ |
bbb77e7c | 144 | ogg_packet *op2= (ogg_packet*)context->buffer; |
6d8f985e | 145 | op2->packet = context->buffer + sizeof(ogg_packet); |
3f4993f1 | 146 | |
6f824977 | 147 | l= op2->bytes; |
42661adf | 148 | avccontext->coded_frame->pts= av_rescale_q(op2->granulepos, (AVRational){1, avccontext->sample_rate}, avccontext->time_base); |
3723f01c | 149 | //FIXME we should reorder the user supplied pts and not assume that they are spaced by 1/sample_rate |
3f4993f1 | 150 | |
6f824977 MN |
151 | memcpy(packets, op2->packet, l); |
152 | context->buffer_index -= l + sizeof(ogg_packet); | |
153 | memcpy(context->buffer, context->buffer + l + sizeof(ogg_packet), context->buffer_index); | |
6d8f985e | 154 | // av_log(avccontext, AV_LOG_DEBUG, "E%d\n", l); |
6d8f985e MN |
155 | } |
156 | ||
3f4993f1 | 157 | return l; |
81e0d0b4 MH |
158 | } |
159 | ||
160 | ||
3aca208a | 161 | static int oggvorbis_encode_close(AVCodecContext *avccontext) { |
81e0d0b4 MH |
162 | OggVorbisContext *context = avccontext->priv_data ; |
163 | /* ogg_packet op ; */ | |
115329f1 | 164 | |
81e0d0b4 MH |
165 | vorbis_analysis_wrote(&context->vd, 0) ; /* notify vorbisenc this is EOF */ |
166 | ||
81e0d0b4 MH |
167 | vorbis_block_clear(&context->vb); |
168 | vorbis_dsp_clear(&context->vd); | |
169 | vorbis_info_clear(&context->vi); | |
492cd3a9 MN |
170 | |
171 | av_freep(&avccontext->coded_frame); | |
bbb77e7c | 172 | av_freep(&avccontext->extradata); |
115329f1 | 173 | |
81e0d0b4 MH |
174 | return 0 ; |
175 | } | |
176 | ||
177 | ||
178 | AVCodec oggvorbis_encoder = { | |
179 | "vorbis", | |
180 | CODEC_TYPE_AUDIO, | |
181 | CODEC_ID_VORBIS, | |
182 | sizeof(OggVorbisContext), | |
183 | oggvorbis_encode_init, | |
184 | oggvorbis_encode_frame, | |
6f824977 MN |
185 | oggvorbis_encode_close, |
186 | .capabilities= CODEC_CAP_DELAY, | |
3aca208a MH |
187 | } ; |
188 | ||
3aca208a MH |
189 | static int oggvorbis_decode_init(AVCodecContext *avccontext) { |
190 | OggVorbisContext *context = avccontext->priv_data ; | |
a03cbe5f | 191 | uint8_t *p= avccontext->extradata; |
ad2b531d MR |
192 | int i, hsizes[3]; |
193 | unsigned char *headers[3], *extradata = avccontext->extradata; | |
3aca208a MH |
194 | |
195 | vorbis_info_init(&context->vi) ; | |
196 | vorbis_comment_init(&context->vc) ; | |
a03cbe5f | 197 | |
ad2b531d MR |
198 | if(! avccontext->extradata_size || ! p) { |
199 | av_log(avccontext, AV_LOG_ERROR, "vorbis extradata absent\n"); | |
200 | return -1; | |
201 | } | |
2c578058 MR |
202 | |
203 | if(p[0] == 0 && p[1] == 30) { | |
204 | for(i = 0; i < 3; i++){ | |
205 | hsizes[i] = *p++ << 8; | |
206 | hsizes[i] += *p++; | |
207 | headers[i] = p; | |
208 | p += hsizes[i]; | |
209 | } | |
210 | } else if(*p == 2) { | |
211 | unsigned int offset = 1; | |
212 | p++; | |
213 | for(i=0; i<2; i++) { | |
214 | hsizes[i] = 0; | |
215 | while((*p == 0xFF) && (offset < avccontext->extradata_size)) { | |
216 | hsizes[i] += 0xFF; | |
217 | offset++; | |
218 | p++; | |
219 | } | |
220 | if(offset >= avccontext->extradata_size - 1) { | |
221 | av_log(avccontext, AV_LOG_ERROR, | |
222 | "vorbis header sizes damaged\n"); | |
223 | return -1; | |
224 | } | |
225 | hsizes[i] += *p; | |
ad2b531d MR |
226 | offset++; |
227 | p++; | |
228 | } | |
2c578058 | 229 | hsizes[2] = avccontext->extradata_size - hsizes[0]-hsizes[1]-offset; |
ad2b531d | 230 | #if 0 |
2c578058 MR |
231 | av_log(avccontext, AV_LOG_DEBUG, |
232 | "vorbis header sizes: %d, %d, %d, / extradata_len is %d \n", | |
233 | hsizes[0], hsizes[1], hsizes[2], avccontext->extradata_size); | |
ad2b531d | 234 | #endif |
2c578058 MR |
235 | headers[0] = extradata + offset; |
236 | headers[1] = extradata + offset + hsizes[0]; | |
237 | headers[2] = extradata + offset + hsizes[0] + hsizes[1]; | |
238 | } else { | |
239 | av_log(avccontext, AV_LOG_ERROR, | |
240 | "vorbis initial header len is wrong: %d\n", *p); | |
241 | return -1; | |
242 | } | |
ad2b531d | 243 | |
a03cbe5f MN |
244 | for(i=0; i<3; i++){ |
245 | context->op.b_o_s= i==0; | |
ad2b531d MR |
246 | context->op.bytes = hsizes[i]; |
247 | context->op.packet = headers[i]; | |
248 | if(vorbis_synthesis_headerin(&context->vi, &context->vc, &context->op)<0){ | |
a03cbe5f MN |
249 | av_log(avccontext, AV_LOG_ERROR, "%d. vorbis header damaged\n", i+1); |
250 | return -1; | |
251 | } | |
252 | } | |
2c578058 | 253 | |
a03cbe5f MN |
254 | avccontext->channels = context->vi.channels; |
255 | avccontext->sample_rate = context->vi.rate; | |
c0df9d75 | 256 | avccontext->time_base= (AVRational){1, avccontext->sample_rate}; |
a03cbe5f MN |
257 | |
258 | vorbis_synthesis_init(&context->vd, &context->vi); | |
115329f1 | 259 | vorbis_block_init(&context->vd, &context->vb); |
3aca208a MH |
260 | |
261 | return 0 ; | |
262 | } | |
263 | ||
264 | ||
265 | static inline int conv(int samples, float **pcm, char *buf, int channels) { | |
266 | int i, j, val ; | |
267 | ogg_int16_t *ptr, *data = (ogg_int16_t*)buf ; | |
268 | float *mono ; | |
115329f1 | 269 | |
3aca208a MH |
270 | for(i = 0 ; i < channels ; i++){ |
271 | ptr = &data[i]; | |
272 | mono = pcm[i] ; | |
115329f1 | 273 | |
3aca208a | 274 | for(j = 0 ; j < samples ; j++) { |
115329f1 | 275 | |
3aca208a | 276 | val = mono[j] * 32767.f; |
115329f1 | 277 | |
3aca208a MH |
278 | if(val > 32767) val = 32767 ; |
279 | if(val < -32768) val = -32768 ; | |
115329f1 | 280 | |
3aca208a MH |
281 | *ptr = val ; |
282 | ptr += channels; | |
283 | } | |
284 | } | |
115329f1 | 285 | |
3aca208a MH |
286 | return 0 ; |
287 | } | |
115329f1 DB |
288 | |
289 | ||
3aca208a MH |
290 | static int oggvorbis_decode_frame(AVCodecContext *avccontext, |
291 | void *data, int *data_size, | |
0c1a9eda | 292 | uint8_t *buf, int buf_size) |
3aca208a MH |
293 | { |
294 | OggVorbisContext *context = avccontext->priv_data ; | |
3aca208a | 295 | float **pcm ; |
115329f1 | 296 | ogg_packet *op= &context->op; |
62ab0b78 | 297 | int samples, total_samples, total_bytes; |
115329f1 | 298 | |
8bfed902 MN |
299 | if(!buf_size){ |
300 | //FIXME flush | |
8bfed902 MN |
301 | return 0; |
302 | } | |
115329f1 | 303 | |
d76f581f MN |
304 | op->packet = buf; |
305 | op->bytes = buf_size; | |
3aca208a | 306 | |
8bfed902 | 307 | // av_log(avccontext, AV_LOG_DEBUG, "%d %d %d %lld %lld %d %d\n", op->bytes, op->b_o_s, op->e_o_s, op->granulepos, op->packetno, buf_size, context->vi.rate); |
115329f1 | 308 | |
8bfed902 MN |
309 | /* for(i=0; i<op->bytes; i++) |
310 | av_log(avccontext, AV_LOG_DEBUG, "%02X ", op->packet[i]); | |
311 | av_log(avccontext, AV_LOG_DEBUG, "\n");*/ | |
3aca208a MH |
312 | |
313 | if(vorbis_synthesis(&context->vb, op) == 0) | |
314 | vorbis_synthesis_blockin(&context->vd, &context->vb) ; | |
115329f1 | 315 | |
3aca208a MH |
316 | total_samples = 0 ; |
317 | total_bytes = 0 ; | |
318 | ||
319 | while((samples = vorbis_synthesis_pcmout(&context->vd, &pcm)) > 0) { | |
320 | conv(samples, pcm, (char*)data + total_bytes, context->vi.channels) ; | |
321 | total_bytes += samples * 2 * context->vi.channels ; | |
322 | total_samples += samples ; | |
323 | vorbis_synthesis_read(&context->vd, samples) ; | |
324 | } | |
325 | ||
115329f1 | 326 | *data_size = total_bytes ; |
3aca208a MH |
327 | return buf_size ; |
328 | } | |
81e0d0b4 | 329 | |
3aca208a MH |
330 | |
331 | static int oggvorbis_decode_close(AVCodecContext *avccontext) { | |
332 | OggVorbisContext *context = avccontext->priv_data ; | |
115329f1 | 333 | |
3aca208a MH |
334 | vorbis_info_clear(&context->vi) ; |
335 | vorbis_comment_clear(&context->vc) ; | |
336 | ||
337 | return 0 ; | |
338 | } | |
339 | ||
340 | ||
341 | AVCodec oggvorbis_decoder = { | |
342 | "vorbis", | |
343 | CODEC_TYPE_AUDIO, | |
344 | CODEC_ID_VORBIS, | |
345 | sizeof(OggVorbisContext), | |
346 | oggvorbis_decode_init, | |
347 | NULL, | |
348 | oggvorbis_decode_close, | |
349 | oggvorbis_decode_frame, | |
6f824977 | 350 | .capabilities= CODEC_CAP_DELAY, |
3aca208a | 351 | } ; |