Commit | Line | Data |
---|---|---|
2ba45a60 DM |
1 | /* |
2 | * Opus encoder using libopus | |
3 | * Copyright (c) 2012 Nathan Caldwell | |
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 | #include <opus.h> | |
23 | #include <opus_multistream.h> | |
24 | ||
25 | #include "libavutil/opt.h" | |
26 | #include "avcodec.h" | |
27 | #include "bytestream.h" | |
28 | #include "internal.h" | |
29 | #include "libopus.h" | |
30 | #include "vorbis.h" | |
31 | #include "audio_frame_queue.h" | |
32 | ||
33 | typedef struct LibopusEncOpts { | |
34 | int vbr; | |
35 | int application; | |
36 | int packet_loss; | |
37 | int complexity; | |
38 | float frame_duration; | |
39 | int packet_size; | |
40 | int max_bandwidth; | |
41 | } LibopusEncOpts; | |
42 | ||
43 | typedef struct LibopusEncContext { | |
44 | AVClass *class; | |
45 | OpusMSEncoder *enc; | |
46 | int stream_count; | |
47 | uint8_t *samples; | |
48 | LibopusEncOpts opts; | |
49 | AudioFrameQueue afq; | |
50 | } LibopusEncContext; | |
51 | ||
52 | static const uint8_t opus_coupled_streams[8] = { | |
53 | 0, 1, 1, 2, 2, 2, 2, 3 | |
54 | }; | |
55 | ||
56 | /* Opus internal to Vorbis channel order mapping written in the header */ | |
57 | static const uint8_t opus_vorbis_channel_map[8][8] = { | |
58 | { 0 }, | |
59 | { 0, 1 }, | |
60 | { 0, 2, 1 }, | |
61 | { 0, 1, 2, 3 }, | |
62 | { 0, 4, 1, 2, 3 }, | |
63 | { 0, 4, 1, 2, 3, 5 }, | |
64 | { 0, 4, 1, 2, 3, 5, 6 }, | |
65 | { 0, 6, 1, 2, 3, 4, 5, 7 }, | |
66 | }; | |
67 | ||
68 | /* libavcodec to libopus channel order mapping, passed to libopus */ | |
69 | static const uint8_t libavcodec_libopus_channel_map[8][8] = { | |
70 | { 0 }, | |
71 | { 0, 1 }, | |
72 | { 0, 1, 2 }, | |
73 | { 0, 1, 2, 3 }, | |
74 | { 0, 1, 3, 4, 2 }, | |
75 | { 0, 1, 4, 5, 2, 3 }, | |
76 | { 0, 1, 5, 6, 2, 4, 3 }, | |
77 | { 0, 1, 6, 7, 4, 5, 2, 3 }, | |
78 | }; | |
79 | ||
80 | static void libopus_write_header(AVCodecContext *avctx, int stream_count, | |
81 | int coupled_stream_count, | |
82 | const uint8_t *channel_mapping) | |
83 | { | |
84 | uint8_t *p = avctx->extradata; | |
85 | int channels = avctx->channels; | |
86 | ||
87 | bytestream_put_buffer(&p, "OpusHead", 8); | |
88 | bytestream_put_byte(&p, 1); /* Version */ | |
89 | bytestream_put_byte(&p, channels); | |
f6fa7814 | 90 | bytestream_put_le16(&p, avctx->initial_padding); /* Lookahead samples at 48kHz */ |
2ba45a60 DM |
91 | bytestream_put_le32(&p, avctx->sample_rate); /* Original sample rate */ |
92 | bytestream_put_le16(&p, 0); /* Gain of 0dB is recommended. */ | |
93 | ||
94 | /* Channel mapping */ | |
95 | if (channels > 2) { | |
96 | bytestream_put_byte(&p, channels <= 8 ? 1 : 255); | |
97 | bytestream_put_byte(&p, stream_count); | |
98 | bytestream_put_byte(&p, coupled_stream_count); | |
99 | bytestream_put_buffer(&p, channel_mapping, channels); | |
100 | } else { | |
101 | bytestream_put_byte(&p, 0); | |
102 | } | |
103 | } | |
104 | ||
105 | static int libopus_configure_encoder(AVCodecContext *avctx, OpusMSEncoder *enc, | |
106 | LibopusEncOpts *opts) | |
107 | { | |
108 | int ret; | |
109 | ||
110 | if (avctx->global_quality) { | |
111 | av_log(avctx, AV_LOG_ERROR, | |
112 | "Quality-based encoding not supported, " | |
113 | "please specify a bitrate and VBR setting.\n"); | |
114 | return AVERROR(EINVAL); | |
115 | } | |
116 | ||
117 | ret = opus_multistream_encoder_ctl(enc, OPUS_SET_BITRATE(avctx->bit_rate)); | |
118 | if (ret != OPUS_OK) { | |
119 | av_log(avctx, AV_LOG_ERROR, | |
120 | "Failed to set bitrate: %s\n", opus_strerror(ret)); | |
121 | return ret; | |
122 | } | |
123 | ||
124 | ret = opus_multistream_encoder_ctl(enc, | |
125 | OPUS_SET_COMPLEXITY(opts->complexity)); | |
126 | if (ret != OPUS_OK) | |
127 | av_log(avctx, AV_LOG_WARNING, | |
128 | "Unable to set complexity: %s\n", opus_strerror(ret)); | |
129 | ||
130 | ret = opus_multistream_encoder_ctl(enc, OPUS_SET_VBR(!!opts->vbr)); | |
131 | if (ret != OPUS_OK) | |
132 | av_log(avctx, AV_LOG_WARNING, | |
133 | "Unable to set VBR: %s\n", opus_strerror(ret)); | |
134 | ||
135 | ret = opus_multistream_encoder_ctl(enc, | |
136 | OPUS_SET_VBR_CONSTRAINT(opts->vbr == 2)); | |
137 | if (ret != OPUS_OK) | |
138 | av_log(avctx, AV_LOG_WARNING, | |
139 | "Unable to set constrained VBR: %s\n", opus_strerror(ret)); | |
140 | ||
141 | ret = opus_multistream_encoder_ctl(enc, | |
142 | OPUS_SET_PACKET_LOSS_PERC(opts->packet_loss)); | |
143 | if (ret != OPUS_OK) | |
144 | av_log(avctx, AV_LOG_WARNING, | |
145 | "Unable to set expected packet loss percentage: %s\n", | |
146 | opus_strerror(ret)); | |
147 | ||
148 | if (avctx->cutoff) { | |
149 | ret = opus_multistream_encoder_ctl(enc, | |
150 | OPUS_SET_MAX_BANDWIDTH(opts->max_bandwidth)); | |
151 | if (ret != OPUS_OK) | |
152 | av_log(avctx, AV_LOG_WARNING, | |
153 | "Unable to set maximum bandwidth: %s\n", opus_strerror(ret)); | |
154 | } | |
155 | ||
156 | return OPUS_OK; | |
157 | } | |
158 | ||
159 | static av_cold int libopus_encode_init(AVCodecContext *avctx) | |
160 | { | |
161 | LibopusEncContext *opus = avctx->priv_data; | |
162 | const uint8_t *channel_mapping; | |
163 | OpusMSEncoder *enc; | |
164 | int ret = OPUS_OK; | |
165 | int coupled_stream_count, header_size, frame_size; | |
166 | ||
167 | coupled_stream_count = opus_coupled_streams[avctx->channels - 1]; | |
168 | opus->stream_count = avctx->channels - coupled_stream_count; | |
169 | channel_mapping = libavcodec_libopus_channel_map[avctx->channels - 1]; | |
170 | ||
171 | /* FIXME: Opus can handle up to 255 channels. However, the mapping for | |
172 | * anything greater than 8 is undefined. */ | |
f6fa7814 DM |
173 | if (avctx->channels > 8) { |
174 | av_log(avctx, AV_LOG_ERROR, | |
2ba45a60 | 175 | "Channel layout undefined for %d channels.\n", avctx->channels); |
f6fa7814 DM |
176 | return AVERROR_PATCHWELCOME; |
177 | } | |
2ba45a60 DM |
178 | if (!avctx->bit_rate) { |
179 | /* Sane default copied from opusenc */ | |
180 | avctx->bit_rate = 64000 * opus->stream_count + | |
181 | 32000 * coupled_stream_count; | |
182 | av_log(avctx, AV_LOG_WARNING, | |
183 | "No bit rate set. Defaulting to %d bps.\n", avctx->bit_rate); | |
184 | } | |
185 | ||
186 | if (avctx->bit_rate < 500 || avctx->bit_rate > 256000 * avctx->channels) { | |
187 | av_log(avctx, AV_LOG_ERROR, "The bit rate %d bps is unsupported. " | |
188 | "Please choose a value between 500 and %d.\n", avctx->bit_rate, | |
189 | 256000 * avctx->channels); | |
190 | return AVERROR(EINVAL); | |
191 | } | |
192 | ||
193 | frame_size = opus->opts.frame_duration * 48000 / 1000; | |
194 | switch (frame_size) { | |
195 | case 120: | |
196 | case 240: | |
197 | if (opus->opts.application != OPUS_APPLICATION_RESTRICTED_LOWDELAY) | |
198 | av_log(avctx, AV_LOG_WARNING, | |
199 | "LPC mode cannot be used with a frame duration of less " | |
200 | "than 10ms. Enabling restricted low-delay mode.\n" | |
201 | "Use a longer frame duration if this is not what you want.\n"); | |
202 | /* Frame sizes less than 10 ms can only use MDCT mode, so switching to | |
203 | * RESTRICTED_LOWDELAY avoids an unnecessary extra 2.5ms lookahead. */ | |
204 | opus->opts.application = OPUS_APPLICATION_RESTRICTED_LOWDELAY; | |
205 | case 480: | |
206 | case 960: | |
207 | case 1920: | |
208 | case 2880: | |
209 | opus->opts.packet_size = | |
210 | avctx->frame_size = frame_size * avctx->sample_rate / 48000; | |
211 | break; | |
212 | default: | |
213 | av_log(avctx, AV_LOG_ERROR, "Invalid frame duration: %g.\n" | |
214 | "Frame duration must be exactly one of: 2.5, 5, 10, 20, 40 or 60.\n", | |
215 | opus->opts.frame_duration); | |
216 | return AVERROR(EINVAL); | |
217 | } | |
218 | ||
219 | if (avctx->compression_level < 0 || avctx->compression_level > 10) { | |
220 | av_log(avctx, AV_LOG_WARNING, | |
221 | "Compression level must be in the range 0 to 10. " | |
222 | "Defaulting to 10.\n"); | |
223 | opus->opts.complexity = 10; | |
224 | } else { | |
225 | opus->opts.complexity = avctx->compression_level; | |
226 | } | |
227 | ||
228 | if (avctx->cutoff) { | |
229 | switch (avctx->cutoff) { | |
230 | case 4000: | |
231 | opus->opts.max_bandwidth = OPUS_BANDWIDTH_NARROWBAND; | |
232 | break; | |
233 | case 6000: | |
234 | opus->opts.max_bandwidth = OPUS_BANDWIDTH_MEDIUMBAND; | |
235 | break; | |
236 | case 8000: | |
237 | opus->opts.max_bandwidth = OPUS_BANDWIDTH_WIDEBAND; | |
238 | break; | |
239 | case 12000: | |
240 | opus->opts.max_bandwidth = OPUS_BANDWIDTH_SUPERWIDEBAND; | |
241 | break; | |
242 | case 20000: | |
243 | opus->opts.max_bandwidth = OPUS_BANDWIDTH_FULLBAND; | |
244 | break; | |
245 | default: | |
246 | av_log(avctx, AV_LOG_WARNING, | |
247 | "Invalid frequency cutoff: %d. Using default maximum bandwidth.\n" | |
248 | "Cutoff frequency must be exactly one of: 4000, 6000, 8000, 12000 or 20000.\n", | |
249 | avctx->cutoff); | |
250 | avctx->cutoff = 0; | |
251 | } | |
252 | } | |
253 | ||
254 | enc = opus_multistream_encoder_create(avctx->sample_rate, avctx->channels, | |
255 | opus->stream_count, | |
256 | coupled_stream_count, | |
257 | channel_mapping, | |
258 | opus->opts.application, &ret); | |
259 | if (ret != OPUS_OK) { | |
260 | av_log(avctx, AV_LOG_ERROR, | |
261 | "Failed to create encoder: %s\n", opus_strerror(ret)); | |
262 | return ff_opus_error_to_averror(ret); | |
263 | } | |
264 | ||
265 | ret = libopus_configure_encoder(avctx, enc, &opus->opts); | |
266 | if (ret != OPUS_OK) { | |
267 | ret = ff_opus_error_to_averror(ret); | |
268 | goto fail; | |
269 | } | |
270 | ||
271 | header_size = 19 + (avctx->channels > 2 ? 2 + avctx->channels : 0); | |
272 | avctx->extradata = av_malloc(header_size + FF_INPUT_BUFFER_PADDING_SIZE); | |
273 | if (!avctx->extradata) { | |
274 | av_log(avctx, AV_LOG_ERROR, "Failed to allocate extradata.\n"); | |
275 | ret = AVERROR(ENOMEM); | |
276 | goto fail; | |
277 | } | |
278 | avctx->extradata_size = header_size; | |
279 | ||
280 | opus->samples = av_mallocz(frame_size * avctx->channels * | |
281 | av_get_bytes_per_sample(avctx->sample_fmt)); | |
282 | if (!opus->samples) { | |
283 | av_log(avctx, AV_LOG_ERROR, "Failed to allocate samples buffer.\n"); | |
284 | ret = AVERROR(ENOMEM); | |
285 | goto fail; | |
286 | } | |
287 | ||
f6fa7814 | 288 | ret = opus_multistream_encoder_ctl(enc, OPUS_GET_LOOKAHEAD(&avctx->initial_padding)); |
2ba45a60 DM |
289 | if (ret != OPUS_OK) |
290 | av_log(avctx, AV_LOG_WARNING, | |
291 | "Unable to get number of lookahead samples: %s\n", | |
292 | opus_strerror(ret)); | |
293 | ||
294 | libopus_write_header(avctx, opus->stream_count, coupled_stream_count, | |
295 | opus_vorbis_channel_map[avctx->channels - 1]); | |
296 | ||
297 | ff_af_queue_init(avctx, &opus->afq); | |
298 | ||
299 | opus->enc = enc; | |
300 | ||
301 | return 0; | |
302 | ||
303 | fail: | |
304 | opus_multistream_encoder_destroy(enc); | |
305 | av_freep(&avctx->extradata); | |
306 | return ret; | |
307 | } | |
308 | ||
309 | static int libopus_encode(AVCodecContext *avctx, AVPacket *avpkt, | |
310 | const AVFrame *frame, int *got_packet_ptr) | |
311 | { | |
312 | LibopusEncContext *opus = avctx->priv_data; | |
313 | const int sample_size = avctx->channels * | |
314 | av_get_bytes_per_sample(avctx->sample_fmt); | |
315 | uint8_t *audio; | |
316 | int ret; | |
317 | int discard_padding; | |
318 | ||
319 | if (frame) { | |
f6fa7814 DM |
320 | ret = ff_af_queue_add(&opus->afq, frame); |
321 | if (ret < 0) | |
322 | return ret; | |
2ba45a60 DM |
323 | if (frame->nb_samples < opus->opts.packet_size) { |
324 | audio = opus->samples; | |
325 | memcpy(audio, frame->data[0], frame->nb_samples * sample_size); | |
326 | } else | |
327 | audio = frame->data[0]; | |
328 | } else { | |
329 | if (!opus->afq.remaining_samples) | |
330 | return 0; | |
331 | audio = opus->samples; | |
332 | memset(audio, 0, opus->opts.packet_size * sample_size); | |
333 | } | |
334 | ||
335 | /* Maximum packet size taken from opusenc in opus-tools. 60ms packets | |
336 | * consist of 3 frames in one packet. The maximum frame size is 1275 | |
337 | * bytes along with the largest possible packet header of 7 bytes. */ | |
338 | if ((ret = ff_alloc_packet2(avctx, avpkt, (1275 * 3 + 7) * opus->stream_count)) < 0) | |
339 | return ret; | |
340 | ||
341 | if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) | |
342 | ret = opus_multistream_encode_float(opus->enc, (float *)audio, | |
343 | opus->opts.packet_size, | |
344 | avpkt->data, avpkt->size); | |
345 | else | |
346 | ret = opus_multistream_encode(opus->enc, (opus_int16 *)audio, | |
347 | opus->opts.packet_size, | |
348 | avpkt->data, avpkt->size); | |
349 | ||
350 | if (ret < 0) { | |
351 | av_log(avctx, AV_LOG_ERROR, | |
352 | "Error encoding frame: %s\n", opus_strerror(ret)); | |
353 | return ff_opus_error_to_averror(ret); | |
354 | } | |
355 | ||
356 | av_shrink_packet(avpkt, ret); | |
357 | ||
358 | ff_af_queue_remove(&opus->afq, opus->opts.packet_size, | |
359 | &avpkt->pts, &avpkt->duration); | |
360 | ||
361 | discard_padding = opus->opts.packet_size - avpkt->duration; | |
362 | // Check if subtraction resulted in an overflow | |
363 | if ((discard_padding < opus->opts.packet_size) != (avpkt->duration > 0)) { | |
364 | av_free_packet(avpkt); | |
365 | av_free(avpkt); | |
366 | return AVERROR(EINVAL); | |
367 | } | |
368 | if (discard_padding > 0) { | |
369 | uint8_t* side_data = av_packet_new_side_data(avpkt, | |
370 | AV_PKT_DATA_SKIP_SAMPLES, | |
371 | 10); | |
372 | if(!side_data) { | |
373 | av_free_packet(avpkt); | |
374 | av_free(avpkt); | |
375 | return AVERROR(ENOMEM); | |
376 | } | |
377 | AV_WL32(side_data + 4, discard_padding); | |
378 | } | |
379 | ||
380 | *got_packet_ptr = 1; | |
381 | ||
382 | return 0; | |
383 | } | |
384 | ||
385 | static av_cold int libopus_encode_close(AVCodecContext *avctx) | |
386 | { | |
387 | LibopusEncContext *opus = avctx->priv_data; | |
388 | ||
389 | opus_multistream_encoder_destroy(opus->enc); | |
390 | ||
391 | ff_af_queue_close(&opus->afq); | |
392 | ||
393 | av_freep(&opus->samples); | |
394 | av_freep(&avctx->extradata); | |
395 | ||
396 | return 0; | |
397 | } | |
398 | ||
399 | #define OFFSET(x) offsetof(LibopusEncContext, opts.x) | |
400 | #define FLAGS AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM | |
401 | static const AVOption libopus_options[] = { | |
402 | { "application", "Intended application type", OFFSET(application), AV_OPT_TYPE_INT, { .i64 = OPUS_APPLICATION_AUDIO }, OPUS_APPLICATION_VOIP, OPUS_APPLICATION_RESTRICTED_LOWDELAY, FLAGS, "application" }, | |
403 | { "voip", "Favor improved speech intelligibility", 0, AV_OPT_TYPE_CONST, { .i64 = OPUS_APPLICATION_VOIP }, 0, 0, FLAGS, "application" }, | |
404 | { "audio", "Favor faithfulness to the input", 0, AV_OPT_TYPE_CONST, { .i64 = OPUS_APPLICATION_AUDIO }, 0, 0, FLAGS, "application" }, | |
405 | { "lowdelay", "Restrict to only the lowest delay modes", 0, AV_OPT_TYPE_CONST, { .i64 = OPUS_APPLICATION_RESTRICTED_LOWDELAY }, 0, 0, FLAGS, "application" }, | |
406 | { "frame_duration", "Duration of a frame in milliseconds", OFFSET(frame_duration), AV_OPT_TYPE_FLOAT, { .dbl = 20.0 }, 2.5, 60.0, FLAGS }, | |
407 | { "packet_loss", "Expected packet loss percentage", OFFSET(packet_loss), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 100, FLAGS }, | |
408 | { "vbr", "Variable bit rate mode", OFFSET(vbr), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 2, FLAGS, "vbr" }, | |
409 | { "off", "Use constant bit rate", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, FLAGS, "vbr" }, | |
410 | { "on", "Use variable bit rate", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, FLAGS, "vbr" }, | |
411 | { "constrained", "Use constrained VBR", 0, AV_OPT_TYPE_CONST, { .i64 = 2 }, 0, 0, FLAGS, "vbr" }, | |
412 | { NULL }, | |
413 | }; | |
414 | ||
415 | static const AVClass libopus_class = { | |
416 | .class_name = "libopus", | |
417 | .item_name = av_default_item_name, | |
418 | .option = libopus_options, | |
419 | .version = LIBAVUTIL_VERSION_INT, | |
420 | }; | |
421 | ||
422 | static const AVCodecDefault libopus_defaults[] = { | |
423 | { "b", "0" }, | |
424 | { "compression_level", "10" }, | |
425 | { NULL }, | |
426 | }; | |
427 | ||
428 | static const int libopus_sample_rates[] = { | |
429 | 48000, 24000, 16000, 12000, 8000, 0, | |
430 | }; | |
431 | ||
432 | AVCodec ff_libopus_encoder = { | |
433 | .name = "libopus", | |
434 | .long_name = NULL_IF_CONFIG_SMALL("libopus Opus"), | |
435 | .type = AVMEDIA_TYPE_AUDIO, | |
436 | .id = AV_CODEC_ID_OPUS, | |
437 | .priv_data_size = sizeof(LibopusEncContext), | |
438 | .init = libopus_encode_init, | |
439 | .encode2 = libopus_encode, | |
440 | .close = libopus_encode_close, | |
441 | .capabilities = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME, | |
442 | .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16, | |
443 | AV_SAMPLE_FMT_FLT, | |
444 | AV_SAMPLE_FMT_NONE }, | |
445 | .channel_layouts = ff_vorbis_channel_layouts, | |
446 | .supported_samplerates = libopus_sample_rates, | |
447 | .priv_class = &libopus_class, | |
448 | .defaults = libopus_defaults, | |
449 | }; |