2 * WMA compatible encoder
3 * Copyright (c) 2007 Michael Niedermayer
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
30 static int encode_init(AVCodecContext * avctx){
31 WMACodecContext *s = avctx->priv_data;
32 int i, flags1, flags2;
37 if(avctx->channels > MAX_CHANNELS) {
38 av_log(avctx, AV_LOG_ERROR, "too many channels: got %i, need %i or fewer",
39 avctx->channels, MAX_CHANNELS);
40 return AVERROR(EINVAL);
43 if (avctx->sample_rate > 48000) {
44 av_log(avctx, AV_LOG_ERROR, "sample rate is too high: %d > 48kHz",
46 return AVERROR(EINVAL);
49 if(avctx->bit_rate < 24*1000) {
50 av_log(avctx, AV_LOG_ERROR, "bitrate too low: got %i, need 24000 or higher\n",
52 return AVERROR(EINVAL);
55 /* extract flag infos */
58 if (avctx->codec->id == AV_CODEC_ID_WMAV1) {
59 extradata= av_malloc(4);
60 avctx->extradata_size= 4;
61 AV_WL16(extradata, flags1);
62 AV_WL16(extradata+2, flags2);
63 } else if (avctx->codec->id == AV_CODEC_ID_WMAV2) {
64 extradata= av_mallocz(10);
65 avctx->extradata_size= 10;
66 AV_WL32(extradata, flags1);
67 AV_WL16(extradata+4, flags2);
70 avctx->extradata= extradata;
71 s->use_exp_vlc = flags2 & 0x0001;
72 s->use_bit_reservoir = flags2 & 0x0002;
73 s->use_variable_block_len = flags2 & 0x0004;
74 if (avctx->channels == 2)
77 ff_wma_init(avctx, flags2);
80 for(i = 0; i < s->nb_block_sizes; i++)
81 ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0, 1.0);
83 s->block_align = avctx->bit_rate * (int64_t)s->frame_len /
84 (avctx->sample_rate * 8);
85 s->block_align = FFMIN(s->block_align, MAX_CODED_SUPERFRAME_SIZE);
86 avctx->block_align = s->block_align;
87 avctx->bit_rate = avctx->block_align * 8LL * avctx->sample_rate /
89 //av_log(NULL, AV_LOG_ERROR, "%d %d %d %d\n", s->block_align, avctx->bit_rate, s->frame_len, avctx->sample_rate);
90 avctx->frame_size = avctx->delay = s->frame_len;
92 #if FF_API_OLD_ENCODE_AUDIO
93 avctx->coded_frame = &s->frame;
94 avcodec_get_frame_defaults(avctx->coded_frame);
101 static void apply_window_and_mdct(AVCodecContext * avctx, const signed short * audio, int len) {
102 WMACodecContext *s = avctx->priv_data;
103 int window_index= s->frame_len_bits - s->block_len_bits;
104 FFTContext *mdct = &s->mdct_ctx[window_index];
106 const float * win = s->windows[window_index];
107 int window_len = 1 << s->block_len_bits;
108 float n = window_len/2;
110 for (channel = 0; channel < avctx->channels; channel++) {
111 memcpy(s->output, s->frame_out[channel], sizeof(float)*window_len);
113 for (i = 0; i < len; i++, j += avctx->channels){
114 s->output[i+window_len] = audio[j] / n * win[window_len - i - 1];
115 s->frame_out[channel][i] = audio[j] / n * win[i];
117 mdct->mdct_calc(mdct, s->coefs[channel], s->output);
121 //FIXME use for decoding too
122 static void init_exp(WMACodecContext *s, int ch, const int *exp_param){
125 float v, *q, max_scale, *q_end;
127 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
128 q = s->exponents[ch];
129 q_end = q + s->block_len;
132 /* XXX: use a table */
133 v = pow(10, *exp_param++ * (1.0 / 16.0));
134 max_scale= FFMAX(max_scale, v);
140 s->max_exponent[ch] = max_scale;
143 static void encode_exp_vlc(WMACodecContext *s, int ch, const int *exp_param){
148 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
149 q = s->exponents[ch];
150 q_end = q + s->block_len;
151 if (s->version == 1) {
152 last_exp= *exp_param++;
153 assert(last_exp-10 >= 0 && last_exp-10 < 32);
154 put_bits(&s->pb, 5, last_exp - 10);
159 int exp = *exp_param++;
160 int code = exp - last_exp + 60;
161 assert(code >= 0 && code < 120);
162 put_bits(&s->pb, ff_aac_scalefactor_bits[code], ff_aac_scalefactor_code[code]);
163 /* XXX: use a table */
169 static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], int total_gain){
170 int v, bsize, ch, coef_nb_bits, parse_exponents;
172 int nb_coefs[MAX_CHANNELS];
173 static const int fixed_exp[25]={20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20};
175 //FIXME remove duplication relative to decoder
176 if (s->use_variable_block_len) {
177 assert(0); //FIXME not implemented
179 /* fixed block len */
180 s->next_block_len_bits = s->frame_len_bits;
181 s->prev_block_len_bits = s->frame_len_bits;
182 s->block_len_bits = s->frame_len_bits;
185 s->block_len = 1 << s->block_len_bits;
186 // assert((s->block_pos + s->block_len) <= s->frame_len);
187 bsize = s->frame_len_bits - s->block_len_bits;
190 v = s->coefs_end[bsize] - s->coefs_start;
191 for(ch = 0; ch < s->nb_channels; ch++)
194 int n4 = s->block_len / 2;
195 mdct_norm = 1.0 / (float)n4;
196 if (s->version == 1) {
197 mdct_norm *= sqrt(n4);
201 if (s->nb_channels == 2) {
202 put_bits(&s->pb, 1, !!s->ms_stereo);
205 for(ch = 0; ch < s->nb_channels; ch++) {
206 s->channel_coded[ch] = 1; //FIXME only set channel_coded when needed, instead of always
207 if (s->channel_coded[ch]) {
208 init_exp(s, ch, fixed_exp);
212 for(ch = 0; ch < s->nb_channels; ch++) {
213 if (s->channel_coded[ch]) {
215 float *coefs, *exponents, mult;
218 coefs1 = s->coefs1[ch];
219 exponents = s->exponents[ch];
220 mult = pow(10, total_gain * 0.05) / s->max_exponent[ch];
222 coefs = src_coefs[ch];
223 if (s->use_noise_coding && 0) {
224 assert(0); //FIXME not implemented
226 coefs += s->coefs_start;
228 for(i = 0;i < n; i++){
229 double t= *coefs++ / (exponents[i] * mult);
230 if(t<-32768 || t>32767)
233 coefs1[i] = lrint(t);
240 for(ch = 0; ch < s->nb_channels; ch++) {
241 int a = s->channel_coded[ch];
242 put_bits(&s->pb, 1, a);
249 for(v= total_gain-1; v>=127; v-= 127)
250 put_bits(&s->pb, 7, 127);
251 put_bits(&s->pb, 7, v);
253 coef_nb_bits= ff_wma_total_gain_to_bits(total_gain);
255 if (s->use_noise_coding) {
256 for(ch = 0; ch < s->nb_channels; ch++) {
257 if (s->channel_coded[ch]) {
259 n = s->exponent_high_sizes[bsize];
261 put_bits(&s->pb, 1, s->high_band_coded[ch][i]= 0);
263 nb_coefs[ch] -= s->exponent_high_bands[bsize][i];
270 if (s->block_len_bits != s->frame_len_bits) {
271 put_bits(&s->pb, 1, parse_exponents);
274 if (parse_exponents) {
275 for(ch = 0; ch < s->nb_channels; ch++) {
276 if (s->channel_coded[ch]) {
277 if (s->use_exp_vlc) {
278 encode_exp_vlc(s, ch, fixed_exp);
280 assert(0); //FIXME not implemented
281 // encode_exp_lsp(s, ch);
286 assert(0); //FIXME not implemented
289 for(ch = 0; ch < s->nb_channels; ch++) {
290 if (s->channel_coded[ch]) {
293 tindex = (ch == 1 && s->ms_stereo);
294 ptr = &s->coefs1[ch][0];
295 eptr = ptr + nb_coefs[ch];
298 for(;ptr < eptr; ptr++){
301 int abs_level= FFABS(level);
303 if(abs_level <= s->coef_vlcs[tindex]->max_level){
304 if(run < s->coef_vlcs[tindex]->levels[abs_level-1])
305 code= run + s->int_table[tindex][abs_level-1];
308 assert(code < s->coef_vlcs[tindex]->n);
309 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], s->coef_vlcs[tindex]->huffcodes[code]);
312 if(1<<coef_nb_bits <= abs_level)
315 put_bits(&s->pb, coef_nb_bits, abs_level);
316 put_bits(&s->pb, s->frame_len_bits, run);
318 put_bits(&s->pb, 1, level < 0); //FIXME the sign is fliped somewhere
325 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], s->coef_vlcs[tindex]->huffcodes[1]);
327 if (s->version == 1 && s->nb_channels >= 2) {
328 avpriv_align_put_bits(&s->pb);
334 static int encode_frame(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], uint8_t *buf, int buf_size, int total_gain){
335 init_put_bits(&s->pb, buf, buf_size);
337 if (s->use_bit_reservoir) {
338 assert(0);//FIXME not implemented
340 if(encode_block(s, src_coefs, total_gain) < 0)
344 avpriv_align_put_bits(&s->pb);
346 return put_bits_count(&s->pb)/8 - s->block_align;
349 static int encode_superframe(AVCodecContext *avctx, AVPacket *avpkt,
350 const AVFrame *frame, int *got_packet_ptr)
352 WMACodecContext *s = avctx->priv_data;
353 const int16_t *samples = (const int16_t *)frame->data[0];
354 int i, total_gain, ret;
356 s->block_len_bits= s->frame_len_bits; //required by non variable block len
357 s->block_len = 1 << s->block_len_bits;
359 apply_window_and_mdct(avctx, samples, frame->nb_samples);
365 for(i = 0; i < s->block_len; i++) {
366 a = s->coefs[0][i]*0.5;
367 b = s->coefs[1][i]*0.5;
368 s->coefs[0][i] = a + b;
369 s->coefs[1][i] = a - b;
373 if ((ret = ff_alloc_packet(avpkt, 2 * MAX_CODED_SUPERFRAME_SIZE))) {
374 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
381 int error = encode_frame(s, s->coefs, avpkt->data, avpkt->size,
388 best = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain);
390 int scoreL = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain - i);
391 int scoreR = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain + i);
392 av_log(NULL, AV_LOG_ERROR, "%d %d %d (%d)\n", scoreL, best, scoreR, total_gain);
393 if(scoreL < FFMIN(best, scoreR)){
396 }else if(scoreR < best){
403 if ((i = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain)) >= 0) {
404 av_log(avctx, AV_LOG_ERROR, "required frame size too large. please "
405 "use a higher bit rate.\n");
406 return AVERROR(EINVAL);
408 assert((put_bits_count(&s->pb) & 7) == 0);
410 put_bits(&s->pb, 8, 'N');
412 flush_put_bits(&s->pb);
414 if (frame->pts != AV_NOPTS_VALUE)
415 avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
417 avpkt->size = s->block_align;
422 AVCodec ff_wmav1_encoder = {
424 .type = AVMEDIA_TYPE_AUDIO,
425 .id = AV_CODEC_ID_WMAV1,
426 .priv_data_size = sizeof(WMACodecContext),
428 .encode2 = encode_superframe,
430 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
431 AV_SAMPLE_FMT_NONE },
432 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"),
435 AVCodec ff_wmav2_encoder = {
437 .type = AVMEDIA_TYPE_AUDIO,
438 .id = AV_CODEC_ID_WMAV2,
439 .priv_data_size = sizeof(WMACodecContext),
441 .encode2 = encode_superframe,
443 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
444 AV_SAMPLE_FMT_NONE },
445 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"),