2 * samplerate conversion for both audio and video
3 * Copyright (c) 2000 Fabrice Bellard
5 * This file is part of FFmpeg.
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.
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.
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
24 * samplerate conversion for both audio and video
30 #include "audioconvert.h"
31 #include "libavutil/opt.h"
32 #include "libavutil/mem.h"
33 #include "libavutil/samplefmt.h"
35 #define MAX_CHANNELS 8
37 struct AVResampleContext;
39 static const char *context_to_name(void *ptr)
41 return "audioresample";
44 static const AVOption options[] = {{NULL}};
45 static const AVClass audioresample_context_class = {
46 "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT
49 struct ReSampleContext {
50 struct AVResampleContext *resample_context;
51 short *temp[MAX_CHANNELS];
55 int input_channels, output_channels, filter_channels;
56 AVAudioConvert *convert_ctx[2];
57 enum AVSampleFormat sample_fmt[2]; ///< input and output sample format
58 unsigned sample_size[2]; ///< size of one sample in sample_fmt
59 short *buffer[2]; ///< buffers used for conversion to S16
60 unsigned buffer_size[2]; ///< sizes of allocated buffers
63 /* n1: number of samples */
64 static void stereo_to_mono(short *output, short *input, int n1)
72 q[0] = (p[0] + p[1]) >> 1;
73 q[1] = (p[2] + p[3]) >> 1;
74 q[2] = (p[4] + p[5]) >> 1;
75 q[3] = (p[6] + p[7]) >> 1;
81 q[0] = (p[0] + p[1]) >> 1;
88 /* n1: number of samples */
89 static void mono_to_stereo(short *output, short *input, int n1)
98 v = p[0]; q[0] = v; q[1] = v;
99 v = p[1]; q[2] = v; q[3] = v;
100 v = p[2]; q[4] = v; q[5] = v;
101 v = p[3]; q[6] = v; q[7] = v;
107 v = p[0]; q[0] = v; q[1] = v;
115 5.1 to stereo input: [fl, fr, c, lfe, rl, rr]
116 - Left = front_left + rear_gain * rear_left + center_gain * center
117 - Right = front_right + rear_gain * rear_right + center_gain * center
118 Where rear_gain is usually around 0.5-1.0 and
119 center_gain is almost always 0.7 (-3 dB)
121 static void surround_to_stereo(short **output, short *input, int channels, int samples)
126 for (i = 0; i < samples; i++) {
135 l = av_clip_int16(fl + (0.5 * rl) + (0.7 * c));
136 r = av_clip_int16(fr + (0.5 * rr) + (0.7 * c));
142 /* increment input. */
147 static void deinterleave(short **output, short *input, int channels, int samples)
151 for (i = 0; i < samples; i++) {
152 for (j = 0; j < channels; j++) {
153 *output[j]++ = *input++;
158 static void interleave(short *output, short **input, int channels, int samples)
162 for (i = 0; i < samples; i++) {
163 for (j = 0; j < channels; j++) {
164 *output++ = *input[j]++;
169 static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
174 for (i = 0; i < n; i++) {
177 *output++ = l; /* left */
178 *output++ = (l / 2) + (r / 2); /* center */
179 *output++ = r; /* right */
180 *output++ = 0; /* left surround */
181 *output++ = 0; /* right surroud */
182 *output++ = 0; /* low freq */
186 #define SUPPORT_RESAMPLE(ch1, ch2, ch3, ch4, ch5, ch6, ch7, ch8) \
187 ch8<<7 | ch7<<6 | ch6<<5 | ch5<<4 | ch4<<3 | ch3<<2 | ch2<<1 | ch1<<0
189 static const uint8_t supported_resampling[MAX_CHANNELS] = {
190 // output ch: 1 2 3 4 5 6 7 8
191 SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 0, 0, 0), // 1 input channel
192 SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 1, 0, 0), // 2 input channels
193 SUPPORT_RESAMPLE(0, 0, 1, 0, 0, 0, 0, 0), // 3 input channels
194 SUPPORT_RESAMPLE(0, 0, 0, 1, 0, 0, 0, 0), // 4 input channels
195 SUPPORT_RESAMPLE(0, 0, 0, 0, 1, 0, 0, 0), // 5 input channels
196 SUPPORT_RESAMPLE(0, 1, 0, 0, 0, 1, 0, 0), // 6 input channels
197 SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 1, 0), // 7 input channels
198 SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 0, 1), // 8 input channels
201 ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
202 int output_rate, int input_rate,
203 enum AVSampleFormat sample_fmt_out,
204 enum AVSampleFormat sample_fmt_in,
205 int filter_length, int log2_phase_count,
206 int linear, double cutoff)
210 if (input_channels > MAX_CHANNELS) {
211 av_log(NULL, AV_LOG_ERROR,
212 "Resampling with input channels greater than %d is unsupported.\n",
216 if (!(supported_resampling[input_channels-1] & (1<<(output_channels-1)))) {
218 av_log(NULL, AV_LOG_ERROR, "Unsupported audio resampling. Allowed "
219 "output channels for %d input channel%s", input_channels,
220 input_channels > 1 ? "s:" : ":");
221 for (i = 0; i < MAX_CHANNELS; i++)
222 if (supported_resampling[input_channels-1] & (1<<i))
223 av_log(NULL, AV_LOG_ERROR, " %d", i + 1);
224 av_log(NULL, AV_LOG_ERROR, "\n");
228 s = av_mallocz(sizeof(ReSampleContext));
230 av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n");
234 s->ratio = (float)output_rate / (float)input_rate;
236 s->input_channels = input_channels;
237 s->output_channels = output_channels;
239 s->filter_channels = s->input_channels;
240 if (s->output_channels < s->filter_channels)
241 s->filter_channels = s->output_channels;
243 s->sample_fmt[0] = sample_fmt_in;
244 s->sample_fmt[1] = sample_fmt_out;
245 s->sample_size[0] = av_get_bytes_per_sample(s->sample_fmt[0]);
246 s->sample_size[1] = av_get_bytes_per_sample(s->sample_fmt[1]);
248 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
249 if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
250 s->sample_fmt[0], 1, NULL, 0))) {
251 av_log(s, AV_LOG_ERROR,
252 "Cannot convert %s sample format to s16 sample format\n",
253 av_get_sample_fmt_name(s->sample_fmt[0]));
259 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
260 if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
261 AV_SAMPLE_FMT_S16, 1, NULL, 0))) {
262 av_log(s, AV_LOG_ERROR,
263 "Cannot convert s16 sample format to %s sample format\n",
264 av_get_sample_fmt_name(s->sample_fmt[1]));
265 av_audio_convert_free(s->convert_ctx[0]);
271 s->resample_context = av_resample_init(output_rate, input_rate,
272 filter_length, log2_phase_count,
275 *(const AVClass**)s->resample_context = &audioresample_context_class;
280 /* resample audio. 'nb_samples' is the number of input samples */
281 /* XXX: optimize it ! */
282 int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
285 short *bufin[MAX_CHANNELS];
286 short *bufout[MAX_CHANNELS];
287 short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS];
288 short *output_bak = NULL;
291 if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) {
293 memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
297 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
298 int istride[1] = { s->sample_size[0] };
299 int ostride[1] = { 2 };
300 const void *ibuf[1] = { input };
302 unsigned input_size = nb_samples * s->input_channels * 2;
304 if (!s->buffer_size[0] || s->buffer_size[0] < input_size) {
305 av_free(s->buffer[0]);
306 s->buffer_size[0] = input_size;
307 s->buffer[0] = av_malloc(s->buffer_size[0]);
309 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
314 obuf[0] = s->buffer[0];
316 if (av_audio_convert(s->convert_ctx[0], obuf, ostride,
317 ibuf, istride, nb_samples * s->input_channels) < 0) {
318 av_log(s->resample_context, AV_LOG_ERROR,
319 "Audio sample format conversion failed\n");
323 input = s->buffer[0];
326 lenout= 2*s->output_channels*nb_samples * s->ratio + 16;
328 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
329 int out_size = lenout * av_get_bytes_per_sample(s->sample_fmt[1]) *
333 if (!s->buffer_size[1] || s->buffer_size[1] < out_size) {
334 av_free(s->buffer[1]);
335 s->buffer_size[1] = out_size;
336 s->buffer[1] = av_malloc(s->buffer_size[1]);
338 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
343 output = s->buffer[1];
346 /* XXX: move those malloc to resample init code */
347 for (i = 0; i < s->filter_channels; i++) {
348 bufin[i] = av_malloc((nb_samples + s->temp_len) * sizeof(short));
349 memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
350 buftmp2[i] = bufin[i] + s->temp_len;
351 bufout[i] = av_malloc(lenout * sizeof(short));
354 if (s->input_channels == 2 && s->output_channels == 1) {
356 stereo_to_mono(buftmp2[0], input, nb_samples);
357 } else if (s->output_channels >= 2 && s->input_channels == 1) {
358 buftmp3[0] = bufout[0];
359 memcpy(buftmp2[0], input, nb_samples * sizeof(short));
360 } else if (s->input_channels == 6 && s->output_channels ==2) {
361 buftmp3[0] = bufout[0];
362 buftmp3[1] = bufout[1];
363 surround_to_stereo(buftmp2, input, s->input_channels, nb_samples);
364 } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) {
365 for (i = 0; i < s->input_channels; i++) {
366 buftmp3[i] = bufout[i];
368 deinterleave(buftmp2, input, s->input_channels, nb_samples);
371 memcpy(buftmp2[0], input, nb_samples * sizeof(short));
374 nb_samples += s->temp_len;
376 /* resample each channel */
377 nb_samples1 = 0; /* avoid warning */
378 for (i = 0; i < s->filter_channels; i++) {
380 int is_last = i + 1 == s->filter_channels;
382 nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i],
383 &consumed, nb_samples, lenout, is_last);
384 s->temp_len = nb_samples - consumed;
385 s->temp[i] = av_realloc(s->temp[i], s->temp_len * sizeof(short));
386 memcpy(s->temp[i], bufin[i] + consumed, s->temp_len * sizeof(short));
389 if (s->output_channels == 2 && s->input_channels == 1) {
390 mono_to_stereo(output, buftmp3[0], nb_samples1);
391 } else if (s->output_channels == 6 && s->input_channels == 2) {
392 ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
393 } else if ((s->output_channels == s->input_channels && s->input_channels >= 2) ||
394 (s->output_channels == 2 && s->input_channels == 6)) {
395 interleave(output, buftmp3, s->output_channels, nb_samples1);
398 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
399 int istride[1] = { 2 };
400 int ostride[1] = { s->sample_size[1] };
401 const void *ibuf[1] = { output };
402 void *obuf[1] = { output_bak };
404 if (av_audio_convert(s->convert_ctx[1], obuf, ostride,
405 ibuf, istride, nb_samples1 * s->output_channels) < 0) {
406 av_log(s->resample_context, AV_LOG_ERROR,
407 "Audio sample format convertion failed\n");
412 for (i = 0; i < s->filter_channels; i++) {
420 void audio_resample_close(ReSampleContext *s)
423 av_resample_close(s->resample_context);
424 for (i = 0; i < s->filter_channels; i++)
425 av_freep(&s->temp[i]);
426 av_freep(&s->buffer[0]);
427 av_freep(&s->buffer[1]);
428 av_audio_convert_free(s->convert_ctx[0]);
429 av_audio_convert_free(s->convert_ctx[1]);