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
28 #include "audioconvert.h"
29 #include "libavutil/opt.h"
30 #include "libavutil/samplefmt.h"
32 #define MAX_CHANNELS 8
34 struct AVResampleContext;
36 static const char *context_to_name(void *ptr)
38 return "audioresample";
41 static const AVOption options[] = {{NULL}};
42 static const AVClass audioresample_context_class = {
43 "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT
46 struct ReSampleContext {
47 struct AVResampleContext *resample_context;
48 short *temp[MAX_CHANNELS];
52 int input_channels, output_channels, filter_channels;
53 AVAudioConvert *convert_ctx[2];
54 enum AVSampleFormat sample_fmt[2]; ///< input and output sample format
55 unsigned sample_size[2]; ///< size of one sample in sample_fmt
56 short *buffer[2]; ///< buffers used for conversion to S16
57 unsigned buffer_size[2]; ///< sizes of allocated buffers
60 /* n1: number of samples */
61 static void stereo_to_mono(short *output, short *input, int n1)
69 q[0] = (p[0] + p[1]) >> 1;
70 q[1] = (p[2] + p[3]) >> 1;
71 q[2] = (p[4] + p[5]) >> 1;
72 q[3] = (p[6] + p[7]) >> 1;
78 q[0] = (p[0] + p[1]) >> 1;
85 /* n1: number of samples */
86 static void mono_to_stereo(short *output, short *input, int n1)
95 v = p[0]; q[0] = v; q[1] = v;
96 v = p[1]; q[2] = v; q[3] = v;
97 v = p[2]; q[4] = v; q[5] = v;
98 v = p[3]; q[6] = v; q[7] = v;
104 v = p[0]; q[0] = v; q[1] = v;
112 5.1 to stereo input: [fl, fr, c, lfe, rl, rr]
113 - Left = front_left + rear_gain * rear_left + center_gain * center
114 - Right = front_right + rear_gain * rear_right + center_gain * center
115 Where rear_gain is usually around 0.5-1.0 and
116 center_gain is almost always 0.7 (-3 dB)
118 static void surround_to_stereo(short **output, short *input, int channels, int samples)
123 for (i = 0; i < samples; i++) {
132 l = av_clip_int16(fl + (0.5 * rl) + (0.7 * c));
133 r = av_clip_int16(fr + (0.5 * rr) + (0.7 * c));
139 /* increment input. */
144 static void deinterleave(short **output, short *input, int channels, int samples)
148 for (i = 0; i < samples; i++) {
149 for (j = 0; j < channels; j++) {
150 *output[j]++ = *input++;
155 static void interleave(short *output, short **input, int channels, int samples)
159 for (i = 0; i < samples; i++) {
160 for (j = 0; j < channels; j++) {
161 *output++ = *input[j]++;
166 static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
171 for (i = 0; i < n; i++) {
174 *output++ = l; /* left */
175 *output++ = (l / 2) + (r / 2); /* center */
176 *output++ = r; /* right */
177 *output++ = 0; /* left surround */
178 *output++ = 0; /* right surroud */
179 *output++ = 0; /* low freq */
183 #define SUPPORT_RESAMPLE(ch1, ch2, ch3, ch4, ch5, ch6, ch7, ch8) \
184 ch8<<7 | ch7<<6 | ch6<<5 | ch5<<4 | ch4<<3 | ch3<<2 | ch2<<1 | ch1<<0
186 static const uint8_t supported_resampling[MAX_CHANNELS] = {
187 // output ch: 1 2 3 4 5 6 7 8
188 SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 0, 0, 0), // 1 input channel
189 SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 1, 0, 0), // 2 input channels
190 SUPPORT_RESAMPLE(0, 0, 1, 0, 0, 0, 0, 0), // 3 input channels
191 SUPPORT_RESAMPLE(0, 0, 0, 1, 0, 0, 0, 0), // 4 input channels
192 SUPPORT_RESAMPLE(0, 0, 0, 0, 1, 0, 0, 0), // 5 input channels
193 SUPPORT_RESAMPLE(0, 1, 0, 0, 0, 1, 0, 0), // 6 input channels
194 SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 1, 0), // 7 input channels
195 SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 0, 1), // 8 input channels
198 ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
199 int output_rate, int input_rate,
200 enum AVSampleFormat sample_fmt_out,
201 enum AVSampleFormat sample_fmt_in,
202 int filter_length, int log2_phase_count,
203 int linear, double cutoff)
207 if (input_channels > MAX_CHANNELS) {
208 av_log(NULL, AV_LOG_ERROR,
209 "Resampling with input channels greater than %d is unsupported.\n",
213 if (!(supported_resampling[input_channels-1] & (1<<(output_channels-1)))) {
215 av_log(NULL, AV_LOG_ERROR, "Unsupported audio resampling. Allowed "
216 "output channels for %d input channel%s", input_channels,
217 input_channels > 1 ? "s:" : ":");
218 for (i = 0; i < MAX_CHANNELS; i++)
219 if (supported_resampling[input_channels-1] & (1<<i))
220 av_log(NULL, AV_LOG_ERROR, " %d", i + 1);
221 av_log(NULL, AV_LOG_ERROR, "\n");
225 s = av_mallocz(sizeof(ReSampleContext));
227 av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n");
231 s->ratio = (float)output_rate / (float)input_rate;
233 s->input_channels = input_channels;
234 s->output_channels = output_channels;
236 s->filter_channels = s->input_channels;
237 if (s->output_channels < s->filter_channels)
238 s->filter_channels = s->output_channels;
240 s->sample_fmt[0] = sample_fmt_in;
241 s->sample_fmt[1] = sample_fmt_out;
242 s->sample_size[0] = av_get_bytes_per_sample(s->sample_fmt[0]);
243 s->sample_size[1] = av_get_bytes_per_sample(s->sample_fmt[1]);
245 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
246 if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
247 s->sample_fmt[0], 1, NULL, 0))) {
248 av_log(s, AV_LOG_ERROR,
249 "Cannot convert %s sample format to s16 sample format\n",
250 av_get_sample_fmt_name(s->sample_fmt[0]));
256 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
257 if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
258 AV_SAMPLE_FMT_S16, 1, NULL, 0))) {
259 av_log(s, AV_LOG_ERROR,
260 "Cannot convert s16 sample format to %s sample format\n",
261 av_get_sample_fmt_name(s->sample_fmt[1]));
262 av_audio_convert_free(s->convert_ctx[0]);
268 s->resample_context = av_resample_init(output_rate, input_rate,
269 filter_length, log2_phase_count,
272 *(const AVClass**)s->resample_context = &audioresample_context_class;
277 /* resample audio. 'nb_samples' is the number of input samples */
278 /* XXX: optimize it ! */
279 int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
282 short *bufin[MAX_CHANNELS];
283 short *bufout[MAX_CHANNELS];
284 short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS];
285 short *output_bak = NULL;
288 if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) {
290 memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
294 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
295 int istride[1] = { s->sample_size[0] };
296 int ostride[1] = { 2 };
297 const void *ibuf[1] = { input };
299 unsigned input_size = nb_samples * s->input_channels * 2;
301 if (!s->buffer_size[0] || s->buffer_size[0] < input_size) {
302 av_free(s->buffer[0]);
303 s->buffer_size[0] = input_size;
304 s->buffer[0] = av_malloc(s->buffer_size[0]);
306 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
311 obuf[0] = s->buffer[0];
313 if (av_audio_convert(s->convert_ctx[0], obuf, ostride,
314 ibuf, istride, nb_samples * s->input_channels) < 0) {
315 av_log(s->resample_context, AV_LOG_ERROR,
316 "Audio sample format conversion failed\n");
320 input = s->buffer[0];
323 lenout= 2*s->output_channels*nb_samples * s->ratio + 16;
325 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
326 int out_size = lenout * av_get_bytes_per_sample(s->sample_fmt[1]) *
330 if (!s->buffer_size[1] || s->buffer_size[1] < out_size) {
331 av_free(s->buffer[1]);
332 s->buffer_size[1] = out_size;
333 s->buffer[1] = av_malloc(s->buffer_size[1]);
335 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
340 output = s->buffer[1];
343 /* XXX: move those malloc to resample init code */
344 for (i = 0; i < s->filter_channels; i++) {
345 bufin[i] = av_malloc((nb_samples + s->temp_len) * sizeof(short));
346 memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
347 buftmp2[i] = bufin[i] + s->temp_len;
348 bufout[i] = av_malloc(lenout * sizeof(short));
351 if (s->input_channels == 2 && s->output_channels == 1) {
353 stereo_to_mono(buftmp2[0], input, nb_samples);
354 } else if (s->output_channels >= 2 && s->input_channels == 1) {
355 buftmp3[0] = bufout[0];
356 memcpy(buftmp2[0], input, nb_samples * sizeof(short));
357 } else if (s->input_channels == 6 && s->output_channels ==2) {
358 buftmp3[0] = bufout[0];
359 buftmp3[1] = bufout[1];
360 surround_to_stereo(buftmp2, input, s->input_channels, nb_samples);
361 } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) {
362 for (i = 0; i < s->input_channels; i++) {
363 buftmp3[i] = bufout[i];
365 deinterleave(buftmp2, input, s->input_channels, nb_samples);
368 memcpy(buftmp2[0], input, nb_samples * sizeof(short));
371 nb_samples += s->temp_len;
373 /* resample each channel */
374 nb_samples1 = 0; /* avoid warning */
375 for (i = 0; i < s->filter_channels; i++) {
377 int is_last = i + 1 == s->filter_channels;
379 nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i],
380 &consumed, nb_samples, lenout, is_last);
381 s->temp_len = nb_samples - consumed;
382 s->temp[i] = av_realloc(s->temp[i], s->temp_len * sizeof(short));
383 memcpy(s->temp[i], bufin[i] + consumed, s->temp_len * sizeof(short));
386 if (s->output_channels == 2 && s->input_channels == 1) {
387 mono_to_stereo(output, buftmp3[0], nb_samples1);
388 } else if (s->output_channels == 6 && s->input_channels == 2) {
389 ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
390 } else if ((s->output_channels == s->input_channels && s->input_channels >= 2) ||
391 (s->output_channels == 2 && s->input_channels == 6)) {
392 interleave(output, buftmp3, s->output_channels, nb_samples1);
395 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
396 int istride[1] = { 2 };
397 int ostride[1] = { s->sample_size[1] };
398 const void *ibuf[1] = { output };
399 void *obuf[1] = { output_bak };
401 if (av_audio_convert(s->convert_ctx[1], obuf, ostride,
402 ibuf, istride, nb_samples1 * s->output_channels) < 0) {
403 av_log(s->resample_context, AV_LOG_ERROR,
404 "Audio sample format convertion failed\n");
409 for (i = 0; i < s->filter_channels; i++) {
417 void audio_resample_close(ReSampleContext *s)
420 av_resample_close(s->resample_context);
421 for (i = 0; i < s->filter_channels; i++)
422 av_freep(&s->temp[i]);
423 av_freep(&s->buffer[0]);
424 av_freep(&s->buffer[1]);
425 av_audio_convert_free(s->convert_ctx[0]);
426 av_audio_convert_free(s->convert_ctx[1]);