2 * WMA compatible encoder
3 * Copyright (c) 2007 Michael Niedermayer
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
22 #include "libavutil/attributes.h"
23 #include "libavutil/internal.h"
28 #include "libavutil/avassert.h"
31 static av_cold int encode_init(AVCodecContext *avctx)
33 WMACodecContext *s = avctx->priv_data;
34 int i, flags1, flags2, block_align;
39 if (avctx->channels > MAX_CHANNELS) {
40 av_log(avctx, AV_LOG_ERROR,
41 "too many channels: got %i, need %i or fewer\n",
42 avctx->channels, MAX_CHANNELS);
43 return AVERROR(EINVAL);
46 if (avctx->sample_rate > 48000) {
47 av_log(avctx, AV_LOG_ERROR, "sample rate is too high: %d > 48kHz\n",
49 return AVERROR(EINVAL);
52 if (avctx->bit_rate < 24 * 1000) {
53 av_log(avctx, AV_LOG_ERROR,
54 "bitrate too low: got %"PRId64", need 24000 or higher\n",
55 (int64_t)avctx->bit_rate);
56 return AVERROR(EINVAL);
59 /* extract flag infos */
62 if (avctx->codec->id == AV_CODEC_ID_WMAV1) {
63 extradata = av_malloc(4);
65 return AVERROR(ENOMEM);
66 avctx->extradata_size = 4;
67 AV_WL16(extradata, flags1);
68 AV_WL16(extradata + 2, flags2);
69 } else if (avctx->codec->id == AV_CODEC_ID_WMAV2) {
70 extradata = av_mallocz(10);
72 return AVERROR(ENOMEM);
73 avctx->extradata_size = 10;
74 AV_WL32(extradata, flags1);
75 AV_WL16(extradata + 4, flags2);
79 avctx->extradata = extradata;
80 s->use_exp_vlc = flags2 & 0x0001;
81 s->use_bit_reservoir = flags2 & 0x0002;
82 s->use_variable_block_len = flags2 & 0x0004;
83 if (avctx->channels == 2)
86 ff_wma_init(avctx, flags2);
89 for (i = 0; i < s->nb_block_sizes; i++)
90 ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0, 1.0);
92 block_align = avctx->bit_rate * (int64_t) s->frame_len /
93 (avctx->sample_rate * 8);
94 block_align = FFMIN(block_align, MAX_CODED_SUPERFRAME_SIZE);
95 avctx->block_align = block_align;
96 avctx->frame_size = avctx->initial_padding = s->frame_len;
101 static void apply_window_and_mdct(AVCodecContext *avctx, const AVFrame *frame)
103 WMACodecContext *s = avctx->priv_data;
104 float **audio = (float **) frame->extended_data;
105 int len = frame->nb_samples;
106 int window_index = s->frame_len_bits - s->block_len_bits;
107 FFTContext *mdct = &s->mdct_ctx[window_index];
109 const float *win = s->windows[window_index];
110 int window_len = 1 << s->block_len_bits;
111 float n = 2.0 * 32768.0 / window_len;
113 for (ch = 0; ch < avctx->channels; ch++) {
114 memcpy(s->output, s->frame_out[ch], window_len * sizeof(*s->output));
115 s->fdsp->vector_fmul_scalar(s->frame_out[ch], audio[ch], n, len);
116 s->fdsp->vector_fmul_reverse(&s->output[window_len], s->frame_out[ch],
118 s->fdsp->vector_fmul(s->frame_out[ch], s->frame_out[ch], win, len);
119 mdct->mdct_calc(mdct, s->coefs[ch], s->output);
123 // FIXME use for decoding too
124 static void init_exp(WMACodecContext *s, int ch, const int *exp_param)
128 float v, *q, max_scale, *q_end;
130 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
131 q = s->exponents[ch];
132 q_end = q + s->block_len;
135 /* XXX: use a table */
136 v = ff_exp10(*exp_param++ *(1.0 / 16.0));
137 max_scale = FFMAX(max_scale, v);
143 s->max_exponent[ch] = max_scale;
146 static void encode_exp_vlc(WMACodecContext *s, int ch, const int *exp_param)
152 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
153 q = s->exponents[ch];
154 q_end = q + s->block_len;
155 if (s->version == 1) {
156 last_exp = *exp_param++;
157 av_assert0(last_exp - 10 >= 0 && last_exp - 10 < 32);
158 put_bits(&s->pb, 5, last_exp - 10);
163 int exp = *exp_param++;
164 int code = exp - last_exp + 60;
165 av_assert1(code >= 0 && code < 120);
166 put_bits(&s->pb, ff_aac_scalefactor_bits[code],
167 ff_aac_scalefactor_code[code]);
168 /* XXX: use a table */
174 static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE],
177 int v, bsize, ch, coef_nb_bits, parse_exponents;
179 int nb_coefs[MAX_CHANNELS];
180 static const int fixed_exp[25] = {
188 // FIXME remove duplication relative to decoder
189 if (s->use_variable_block_len) {
190 av_assert0(0); // FIXME not implemented
192 /* fixed block len */
193 s->next_block_len_bits = s->frame_len_bits;
194 s->prev_block_len_bits = s->frame_len_bits;
195 s->block_len_bits = s->frame_len_bits;
198 s->block_len = 1 << s->block_len_bits;
199 // av_assert0((s->block_pos + s->block_len) <= s->frame_len);
200 bsize = s->frame_len_bits - s->block_len_bits;
203 v = s->coefs_end[bsize] - s->coefs_start;
204 for (ch = 0; ch < s->avctx->channels; ch++)
207 int n4 = s->block_len / 2;
208 mdct_norm = 1.0 / (float) n4;
210 mdct_norm *= sqrt(n4);
213 if (s->avctx->channels == 2)
214 put_bits(&s->pb, 1, !!s->ms_stereo);
216 for (ch = 0; ch < s->avctx->channels; ch++) {
217 // FIXME only set channel_coded when needed, instead of always
218 s->channel_coded[ch] = 1;
219 if (s->channel_coded[ch])
220 init_exp(s, ch, fixed_exp);
223 for (ch = 0; ch < s->avctx->channels; ch++) {
224 if (s->channel_coded[ch]) {
226 float *coefs, *exponents, mult;
229 coefs1 = s->coefs1[ch];
230 exponents = s->exponents[ch];
231 mult = ff_exp10(total_gain * 0.05) / s->max_exponent[ch];
233 coefs = src_coefs[ch];
234 if (s->use_noise_coding && 0) {
235 av_assert0(0); // FIXME not implemented
237 coefs += s->coefs_start;
239 for (i = 0; i < n; i++) {
240 double t = *coefs++ / (exponents[i] * mult);
241 if (t < -32768 || t > 32767)
244 coefs1[i] = lrint(t);
251 for (ch = 0; ch < s->avctx->channels; ch++) {
252 int a = s->channel_coded[ch];
253 put_bits(&s->pb, 1, a);
260 for (v = total_gain - 1; v >= 127; v -= 127)
261 put_bits(&s->pb, 7, 127);
262 put_bits(&s->pb, 7, v);
264 coef_nb_bits = ff_wma_total_gain_to_bits(total_gain);
266 if (s->use_noise_coding) {
267 for (ch = 0; ch < s->avctx->channels; ch++) {
268 if (s->channel_coded[ch]) {
270 n = s->exponent_high_sizes[bsize];
271 for (i = 0; i < n; i++) {
272 put_bits(&s->pb, 1, s->high_band_coded[ch][i] = 0);
274 nb_coefs[ch] -= s->exponent_high_bands[bsize][i];
281 if (s->block_len_bits != s->frame_len_bits)
282 put_bits(&s->pb, 1, parse_exponents);
284 if (parse_exponents) {
285 for (ch = 0; ch < s->avctx->channels; ch++) {
286 if (s->channel_coded[ch]) {
287 if (s->use_exp_vlc) {
288 encode_exp_vlc(s, ch, fixed_exp);
290 av_assert0(0); // FIXME not implemented
291 // encode_exp_lsp(s, ch);
296 av_assert0(0); // FIXME not implemented
298 for (ch = 0; ch < s->avctx->channels; ch++) {
299 if (s->channel_coded[ch]) {
302 tindex = (ch == 1 && s->ms_stereo);
303 ptr = &s->coefs1[ch][0];
304 eptr = ptr + nb_coefs[ch];
307 for (; ptr < eptr; ptr++) {
310 int abs_level = FFABS(level);
312 if (abs_level <= s->coef_vlcs[tindex]->max_level)
313 if (run < s->coef_vlcs[tindex]->levels[abs_level - 1])
314 code = run + s->int_table[tindex][abs_level - 1];
316 av_assert2(code < s->coef_vlcs[tindex]->n);
317 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code],
318 s->coef_vlcs[tindex]->huffcodes[code]);
321 if (1 << coef_nb_bits <= abs_level)
324 put_bits(&s->pb, coef_nb_bits, abs_level);
325 put_bits(&s->pb, s->frame_len_bits, run);
327 // FIXME the sign is flipped somewhere
328 put_bits(&s->pb, 1, level < 0);
334 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1],
335 s->coef_vlcs[tindex]->huffcodes[1]);
337 if (s->version == 1 && s->avctx->channels >= 2)
338 avpriv_align_put_bits(&s->pb);
343 static int encode_frame(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE],
344 uint8_t *buf, int buf_size, int total_gain)
346 init_put_bits(&s->pb, buf, buf_size);
348 if (s->use_bit_reservoir)
349 av_assert0(0); // FIXME not implemented
350 else if (encode_block(s, src_coefs, total_gain) < 0)
353 avpriv_align_put_bits(&s->pb);
355 return put_bits_count(&s->pb) / 8 - s->avctx->block_align;
358 static int encode_superframe(AVCodecContext *avctx, AVPacket *avpkt,
359 const AVFrame *frame, int *got_packet_ptr)
361 WMACodecContext *s = avctx->priv_data;
362 int i, total_gain, ret, error;
364 s->block_len_bits = s->frame_len_bits; // required by non variable block len
365 s->block_len = 1 << s->block_len_bits;
367 apply_window_and_mdct(avctx, frame);
373 for (i = 0; i < s->block_len; i++) {
374 a = s->coefs[0][i] * 0.5;
375 b = s->coefs[1][i] * 0.5;
376 s->coefs[0][i] = a + b;
377 s->coefs[1][i] = a - b;
381 if ((ret = ff_alloc_packet2(avctx, avpkt, 2 * MAX_CODED_SUPERFRAME_SIZE, 0)) < 0)
385 for (i = 64; i; i >>= 1) {
386 error = encode_frame(s, s->coefs, avpkt->data, avpkt->size,
392 while(total_gain <= 128 && error > 0)
393 error = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain++);
395 av_log(avctx, AV_LOG_ERROR, "Invalid input data or requested bitrate too low, cannot encode\n");
397 return AVERROR(EINVAL);
399 av_assert0((put_bits_count(&s->pb) & 7) == 0);
400 i= avctx->block_align - (put_bits_count(&s->pb)+7)/8;
403 put_bits(&s->pb, 8, 'N');
405 flush_put_bits(&s->pb);
406 av_assert0(put_bits_ptr(&s->pb) - s->pb.buf == avctx->block_align);
408 if (frame->pts != AV_NOPTS_VALUE)
409 avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->initial_padding);
411 avpkt->size = avctx->block_align;
416 #if CONFIG_WMAV1_ENCODER
417 AVCodec ff_wmav1_encoder = {
419 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"),
420 .type = AVMEDIA_TYPE_AUDIO,
421 .id = AV_CODEC_ID_WMAV1,
422 .priv_data_size = sizeof(WMACodecContext),
424 .encode2 = encode_superframe,
426 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
427 AV_SAMPLE_FMT_NONE },
430 #if CONFIG_WMAV2_ENCODER
431 AVCodec ff_wmav2_encoder = {
433 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"),
434 .type = AVMEDIA_TYPE_AUDIO,
435 .id = AV_CODEC_ID_WMAV2,
436 .priv_data_size = sizeof(WMACodecContext),
438 .encode2 = encode_superframe,
440 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLTP,
441 AV_SAMPLE_FMT_NONE },