blob: 4bdb9618dd3c47cd363eb81a6954dfd3640e4c32
1 | /* |
2 | * Copyright (C) 2009 Justin Ruggles |
3 | * Copyright (c) 2009 Xuggle Incorporated |
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 | * libspeex Speex audio encoder |
25 | * |
26 | * Usage Guide |
27 | * This explains the values that need to be set prior to initialization in |
28 | * order to control various encoding parameters. |
29 | * |
30 | * Channels |
31 | * Speex only supports mono or stereo, so avctx->channels must be set to |
32 | * 1 or 2. |
33 | * |
34 | * Sample Rate / Encoding Mode |
35 | * Speex has 3 modes, each of which uses a specific sample rate. |
36 | * narrowband : 8 kHz |
37 | * wideband : 16 kHz |
38 | * ultra-wideband : 32 kHz |
39 | * avctx->sample_rate must be set to one of these 3 values. This will be |
40 | * used to set the encoding mode. |
41 | * |
42 | * Rate Control |
43 | * VBR mode is turned on by setting AV_CODEC_FLAG_QSCALE in avctx->flags. |
44 | * avctx->global_quality is used to set the encoding quality. |
45 | * For CBR mode, avctx->bit_rate can be used to set the constant bitrate. |
46 | * Alternatively, the 'cbr_quality' option can be set from 0 to 10 to set |
47 | * a constant bitrate based on quality. |
48 | * For ABR mode, set avctx->bit_rate and set the 'abr' option to 1. |
49 | * Approx. Bitrate Range: |
50 | * narrowband : 2400 - 25600 bps |
51 | * wideband : 4000 - 43200 bps |
52 | * ultra-wideband : 4400 - 45200 bps |
53 | * |
54 | * Complexity |
55 | * Encoding complexity is controlled by setting avctx->compression_level. |
56 | * The valid range is 0 to 10. A higher setting gives generally better |
57 | * quality at the expense of encoding speed. This does not affect the |
58 | * bit rate. |
59 | * |
60 | * Frames-per-Packet |
61 | * The encoder defaults to using 1 frame-per-packet. However, it is |
62 | * sometimes desirable to use multiple frames-per-packet to reduce the |
63 | * amount of container overhead. This can be done by setting the |
64 | * 'frames_per_packet' option to a value 1 to 8. |
65 | * |
66 | * |
67 | * Optional features |
68 | * Speex encoder supports several optional features, which can be useful |
69 | * for some conditions. |
70 | * |
71 | * Voice Activity Detection |
72 | * When enabled, voice activity detection detects whether the audio |
73 | * being encoded is speech or silence/background noise. VAD is always |
74 | * implicitly activated when encoding in VBR, so the option is only useful |
75 | * in non-VBR operation. In this case, Speex detects non-speech periods and |
76 | * encodes them with just enough bits to reproduce the background noise. |
77 | * |
78 | * Discontinuous Transmission (DTX) |
79 | * DTX is an addition to VAD/VBR operation, that makes it possible to stop transmitting |
80 | * completely when the background noise is stationary. |
81 | * In file-based operation only 5 bits are used for such frames. |
82 | */ |
83 | |
84 | #include <speex/speex.h> |
85 | #include <speex/speex_header.h> |
86 | #include <speex/speex_stereo.h> |
87 | |
88 | #include "libavutil/channel_layout.h" |
89 | #include "libavutil/common.h" |
90 | #include "libavutil/opt.h" |
91 | #include "avcodec.h" |
92 | #include "internal.h" |
93 | #include "audio_frame_queue.h" |
94 | |
95 | /* TODO: Think about converting abr, vad, dtx and such flags to a bit field */ |
96 | typedef struct LibSpeexEncContext { |
97 | AVClass *class; ///< AVClass for private options |
98 | SpeexBits bits; ///< libspeex bitwriter context |
99 | SpeexHeader header; ///< libspeex header struct |
100 | void *enc_state; ///< libspeex encoder state |
101 | int frames_per_packet; ///< number of frames to encode in each packet |
102 | float vbr_quality; ///< VBR quality 0.0 to 10.0 |
103 | int cbr_quality; ///< CBR quality 0 to 10 |
104 | int abr; ///< flag to enable ABR |
105 | int vad; ///< flag to enable VAD |
106 | int dtx; ///< flag to enable DTX |
107 | int pkt_frame_count; ///< frame count for the current packet |
108 | AudioFrameQueue afq; ///< frame queue |
109 | } LibSpeexEncContext; |
110 | |
111 | static av_cold void print_enc_params(AVCodecContext *avctx, |
112 | LibSpeexEncContext *s) |
113 | { |
114 | const char *mode_str = "unknown"; |
115 | |
116 | av_log(avctx, AV_LOG_DEBUG, "channels: %d\n", avctx->channels); |
117 | switch (s->header.mode) { |
118 | case SPEEX_MODEID_NB: mode_str = "narrowband"; break; |
119 | case SPEEX_MODEID_WB: mode_str = "wideband"; break; |
120 | case SPEEX_MODEID_UWB: mode_str = "ultra-wideband"; break; |
121 | } |
122 | av_log(avctx, AV_LOG_DEBUG, "mode: %s\n", mode_str); |
123 | if (s->header.vbr) { |
124 | av_log(avctx, AV_LOG_DEBUG, "rate control: VBR\n"); |
125 | av_log(avctx, AV_LOG_DEBUG, " quality: %f\n", s->vbr_quality); |
126 | } else if (s->abr) { |
127 | av_log(avctx, AV_LOG_DEBUG, "rate control: ABR\n"); |
128 | av_log(avctx, AV_LOG_DEBUG, " bitrate: %"PRId64" bps\n", (int64_t)avctx->bit_rate); |
129 | } else { |
130 | av_log(avctx, AV_LOG_DEBUG, "rate control: CBR\n"); |
131 | av_log(avctx, AV_LOG_DEBUG, " bitrate: %"PRId64" bps\n", (int64_t)avctx->bit_rate); |
132 | } |
133 | av_log(avctx, AV_LOG_DEBUG, "complexity: %d\n", |
134 | avctx->compression_level); |
135 | av_log(avctx, AV_LOG_DEBUG, "frame size: %d samples\n", |
136 | avctx->frame_size); |
137 | av_log(avctx, AV_LOG_DEBUG, "frames per packet: %d\n", |
138 | s->frames_per_packet); |
139 | av_log(avctx, AV_LOG_DEBUG, "packet size: %d\n", |
140 | avctx->frame_size * s->frames_per_packet); |
141 | av_log(avctx, AV_LOG_DEBUG, "voice activity detection: %d\n", s->vad); |
142 | av_log(avctx, AV_LOG_DEBUG, "discontinuous transmission: %d\n", s->dtx); |
143 | } |
144 | |
145 | static av_cold int encode_init(AVCodecContext *avctx) |
146 | { |
147 | LibSpeexEncContext *s = avctx->priv_data; |
148 | const SpeexMode *mode; |
149 | uint8_t *header_data; |
150 | int header_size; |
151 | int32_t complexity; |
152 | |
153 | /* channels */ |
154 | if (avctx->channels < 1 || avctx->channels > 2) { |
155 | av_log(avctx, AV_LOG_ERROR, "Invalid channels (%d). Only stereo and " |
156 | "mono are supported\n", avctx->channels); |
157 | return AVERROR(EINVAL); |
158 | } |
159 | |
160 | /* sample rate and encoding mode */ |
161 | switch (avctx->sample_rate) { |
162 | case 8000: mode = &speex_nb_mode; break; |
163 | case 16000: mode = &speex_wb_mode; break; |
164 | case 32000: mode = &speex_uwb_mode; break; |
165 | default: |
166 | av_log(avctx, AV_LOG_ERROR, "Sample rate of %d Hz is not supported. " |
167 | "Resample to 8, 16, or 32 kHz.\n", avctx->sample_rate); |
168 | return AVERROR(EINVAL); |
169 | } |
170 | |
171 | /* initialize libspeex */ |
172 | s->enc_state = speex_encoder_init(mode); |
173 | if (!s->enc_state) { |
174 | av_log(avctx, AV_LOG_ERROR, "Error initializing libspeex\n"); |
175 | return -1; |
176 | } |
177 | speex_init_header(&s->header, avctx->sample_rate, avctx->channels, mode); |
178 | |
179 | /* rate control method and parameters */ |
180 | if (avctx->flags & AV_CODEC_FLAG_QSCALE) { |
181 | /* VBR */ |
182 | s->header.vbr = 1; |
183 | s->vad = 1; /* VAD is always implicitly activated for VBR */ |
184 | speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR, &s->header.vbr); |
185 | s->vbr_quality = av_clipf(avctx->global_quality / (float)FF_QP2LAMBDA, |
186 | 0.0f, 10.0f); |
187 | speex_encoder_ctl(s->enc_state, SPEEX_SET_VBR_QUALITY, &s->vbr_quality); |
188 | } else { |
189 | s->header.bitrate = avctx->bit_rate; |
190 | if (avctx->bit_rate > 0) { |
191 | /* CBR or ABR by bitrate */ |
192 | if (s->abr) { |
193 | speex_encoder_ctl(s->enc_state, SPEEX_SET_ABR, |
194 | &s->header.bitrate); |
195 | speex_encoder_ctl(s->enc_state, SPEEX_GET_ABR, |
196 | &s->header.bitrate); |
197 | } else { |
198 | speex_encoder_ctl(s->enc_state, SPEEX_SET_BITRATE, |
199 | &s->header.bitrate); |
200 | speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE, |
201 | &s->header.bitrate); |
202 | } |
203 | } else { |
204 | /* CBR by quality */ |
205 | speex_encoder_ctl(s->enc_state, SPEEX_SET_QUALITY, |
206 | &s->cbr_quality); |
207 | speex_encoder_ctl(s->enc_state, SPEEX_GET_BITRATE, |
208 | &s->header.bitrate); |
209 | } |
210 | /* stereo side information adds about 800 bps to the base bitrate */ |
211 | /* TODO: this should be calculated exactly */ |
212 | avctx->bit_rate = s->header.bitrate + (avctx->channels == 2 ? 800 : 0); |
213 | } |
214 | |
215 | /* VAD is activated with VBR or can be turned on by itself */ |
216 | if (s->vad) |
217 | speex_encoder_ctl(s->enc_state, SPEEX_SET_VAD, &s->vad); |
218 | |
219 | /* Activating Discontinuous Transmission */ |
220 | if (s->dtx) { |
221 | speex_encoder_ctl(s->enc_state, SPEEX_SET_DTX, &s->dtx); |
222 | if (!(s->abr || s->vad || s->header.vbr)) |
223 | av_log(avctx, AV_LOG_WARNING, "DTX is not much of use without ABR, VAD or VBR\n"); |
224 | } |
225 | |
226 | /* set encoding complexity */ |
227 | if (avctx->compression_level > FF_COMPRESSION_DEFAULT) { |
228 | complexity = av_clip(avctx->compression_level, 0, 10); |
229 | speex_encoder_ctl(s->enc_state, SPEEX_SET_COMPLEXITY, &complexity); |
230 | } |
231 | speex_encoder_ctl(s->enc_state, SPEEX_GET_COMPLEXITY, &complexity); |
232 | avctx->compression_level = complexity; |
233 | |
234 | /* set packet size */ |
235 | avctx->frame_size = s->header.frame_size; |
236 | s->header.frames_per_packet = s->frames_per_packet; |
237 | |
238 | /* set encoding delay */ |
239 | speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &avctx->initial_padding); |
240 | ff_af_queue_init(avctx, &s->afq); |
241 | |
242 | /* create header packet bytes from header struct */ |
243 | /* note: libspeex allocates the memory for header_data, which is freed |
244 | below with speex_header_free() */ |
245 | header_data = speex_header_to_packet(&s->header, &header_size); |
246 | |
247 | /* allocate extradata */ |
248 | avctx->extradata = av_malloc(header_size + AV_INPUT_BUFFER_PADDING_SIZE); |
249 | if (!avctx->extradata) { |
250 | speex_header_free(header_data); |
251 | speex_encoder_destroy(s->enc_state); |
252 | av_log(avctx, AV_LOG_ERROR, "memory allocation error\n"); |
253 | return AVERROR(ENOMEM); |
254 | } |
255 | |
256 | /* copy header packet to extradata */ |
257 | memcpy(avctx->extradata, header_data, header_size); |
258 | avctx->extradata_size = header_size; |
259 | speex_header_free(header_data); |
260 | |
261 | /* init libspeex bitwriter */ |
262 | speex_bits_init(&s->bits); |
263 | |
264 | print_enc_params(avctx, s); |
265 | return 0; |
266 | } |
267 | |
268 | static int encode_frame(AVCodecContext *avctx, AVPacket *avpkt, |
269 | const AVFrame *frame, int *got_packet_ptr) |
270 | { |
271 | LibSpeexEncContext *s = avctx->priv_data; |
272 | int16_t *samples = frame ? (int16_t *)frame->data[0] : NULL; |
273 | int ret; |
274 | |
275 | if (samples) { |
276 | /* encode Speex frame */ |
277 | if (avctx->channels == 2) |
278 | speex_encode_stereo_int(samples, s->header.frame_size, &s->bits); |
279 | speex_encode_int(s->enc_state, samples, &s->bits); |
280 | s->pkt_frame_count++; |
281 | if ((ret = ff_af_queue_add(&s->afq, frame)) < 0) |
282 | return ret; |
283 | } else { |
284 | /* handle end-of-stream */ |
285 | if (!s->pkt_frame_count) |
286 | return 0; |
287 | /* add extra terminator codes for unused frames in last packet */ |
288 | while (s->pkt_frame_count < s->frames_per_packet) { |
289 | speex_bits_pack(&s->bits, 15, 5); |
290 | s->pkt_frame_count++; |
291 | } |
292 | } |
293 | |
294 | /* write output if all frames for the packet have been encoded */ |
295 | if (s->pkt_frame_count == s->frames_per_packet) { |
296 | s->pkt_frame_count = 0; |
297 | if ((ret = ff_alloc_packet2(avctx, avpkt, speex_bits_nbytes(&s->bits), 0)) < 0) |
298 | return ret; |
299 | ret = speex_bits_write(&s->bits, avpkt->data, avpkt->size); |
300 | speex_bits_reset(&s->bits); |
301 | |
302 | /* Get the next frame pts/duration */ |
303 | ff_af_queue_remove(&s->afq, s->frames_per_packet * avctx->frame_size, |
304 | &avpkt->pts, &avpkt->duration); |
305 | |
306 | avpkt->size = ret; |
307 | *got_packet_ptr = 1; |
308 | return 0; |
309 | } |
310 | return 0; |
311 | } |
312 | |
313 | static av_cold int encode_close(AVCodecContext *avctx) |
314 | { |
315 | LibSpeexEncContext *s = avctx->priv_data; |
316 | |
317 | speex_bits_destroy(&s->bits); |
318 | speex_encoder_destroy(s->enc_state); |
319 | |
320 | ff_af_queue_close(&s->afq); |
321 | av_freep(&avctx->extradata); |
322 | |
323 | return 0; |
324 | } |
325 | |
326 | #define OFFSET(x) offsetof(LibSpeexEncContext, x) |
327 | #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM |
328 | static const AVOption options[] = { |
329 | { "abr", "Use average bit rate", OFFSET(abr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE }, |
330 | { "cbr_quality", "Set quality value (0 to 10) for CBR", OFFSET(cbr_quality), AV_OPT_TYPE_INT, { .i64 = 8 }, 0, 10, AE }, |
331 | { "frames_per_packet", "Number of frames to encode in each packet", OFFSET(frames_per_packet), AV_OPT_TYPE_INT, { .i64 = 1 }, 1, 8, AE }, |
332 | { "vad", "Voice Activity Detection", OFFSET(vad), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE }, |
333 | { "dtx", "Discontinuous Transmission", OFFSET(dtx), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE }, |
334 | { NULL }, |
335 | }; |
336 | |
337 | static const AVClass speex_class = { |
338 | .class_name = "libspeex", |
339 | .item_name = av_default_item_name, |
340 | .option = options, |
341 | .version = LIBAVUTIL_VERSION_INT, |
342 | }; |
343 | |
344 | static const AVCodecDefault defaults[] = { |
345 | { "b", "0" }, |
346 | { "compression_level", "3" }, |
347 | { NULL }, |
348 | }; |
349 | |
350 | AVCodec ff_libspeex_encoder = { |
351 | .name = "libspeex", |
352 | .long_name = NULL_IF_CONFIG_SMALL("libspeex Speex"), |
353 | .type = AVMEDIA_TYPE_AUDIO, |
354 | .id = AV_CODEC_ID_SPEEX, |
355 | .priv_data_size = sizeof(LibSpeexEncContext), |
356 | .init = encode_init, |
357 | .encode2 = encode_frame, |
358 | .close = encode_close, |
359 | .capabilities = AV_CODEC_CAP_DELAY, |
360 | .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16, |
361 | AV_SAMPLE_FMT_NONE }, |
362 | .channel_layouts = (const uint64_t[]){ AV_CH_LAYOUT_MONO, |
363 | AV_CH_LAYOUT_STEREO, |
364 | 0 }, |
365 | .supported_samplerates = (const int[]){ 8000, 16000, 32000, 0 }, |
366 | .priv_class = &speex_class, |
367 | .defaults = defaults, |
368 | }; |
369 |