1 /*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 #include <rubberband/rubberband-c.h>
20
21 #include "libavutil/channel_layout.h"
22 #include "libavutil/common.h"
23 #include "libavutil/opt.h"
24
25 #include "audio.h"
26 #include "avfilter.h"
27 #include "filters.h"
28 #include "formats.h"
29 #include "internal.h"
30
31 typedef struct RubberBandContext {
32 const AVClass *class;
33 RubberBandState rbs;
34
35 double tempo, pitch;
36 int transients, detector, phase, window,
37 smoothing, formant, opitch, channels;
38 int64_t nb_samples_out;
39 int64_t nb_samples_in;
40 int64_t first_pts;
41 int nb_samples;
42 } RubberBandContext;
43
44 #define OFFSET(x) offsetof(RubberBandContext, x)
45 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
46 #define AT AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
47
48 static const AVOption rubberband_options[] = {
49 { "tempo", "set tempo scale factor", OFFSET(tempo), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.01, 100, AT },
50 { "pitch", "set pitch scale factor", OFFSET(pitch), AV_OPT_TYPE_DOUBLE, {.dbl=1}, 0.01, 100, AT },
51 { "transients", "set transients", OFFSET(transients), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "transients" },
52 { "crisp", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionTransientsCrisp}, 0, 0, A, "transients" },
53 { "mixed", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionTransientsMixed}, 0, 0, A, "transients" },
54 { "smooth", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionTransientsSmooth}, 0, 0, A, "transients" },
55 { "detector", "set detector", OFFSET(detector), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "detector" },
56 { "compound", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionDetectorCompound}, 0, 0, A, "detector" },
57 { "percussive", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionDetectorPercussive}, 0, 0, A, "detector" },
58 { "soft", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionDetectorSoft}, 0, 0, A, "detector" },
59 { "phase", "set phase", OFFSET(phase), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "phase" },
60 { "laminar", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionPhaseLaminar}, 0, 0, A, "phase" },
61 { "independent", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionPhaseIndependent}, 0, 0, A, "phase" },
62 { "window", "set window", OFFSET(window), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "window" },
63 { "standard", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionWindowStandard}, 0, 0, A, "window" },
64 { "short", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionWindowShort}, 0, 0, A, "window" },
65 { "long", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionWindowLong}, 0, 0, A, "window" },
66 { "smoothing", "set smoothing", OFFSET(smoothing), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "smoothing" },
67 { "off", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionSmoothingOff}, 0, 0, A, "smoothing" },
68 { "on", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionSmoothingOn}, 0, 0, A, "smoothing" },
69 { "formant", "set formant", OFFSET(formant), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "formant" },
70 { "shifted", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionFormantShifted}, 0, 0, A, "formant" },
71 { "preserved", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionFormantPreserved}, 0, 0, A, "formant" },
72 { "pitchq", "set pitch quality", OFFSET(opitch), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "pitch" },
73 { "quality", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionPitchHighQuality}, 0, 0, A, "pitch" },
74 { "speed", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionPitchHighSpeed}, 0, 0, A, "pitch" },
75 { "consistency", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionPitchHighConsistency}, 0, 0, A, "pitch" },
76 { "channels", "set channels", OFFSET(channels), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, A, "channels" },
77 { "apart", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionChannelsApart}, 0, 0, A, "channels" },
78 { "together", 0, 0, AV_OPT_TYPE_CONST, {.i64=RubberBandOptionChannelsTogether}, 0, 0, A, "channels" },
79 { NULL },
80 };
81
82 AVFILTER_DEFINE_CLASS(rubberband);
83
uninit(AVFilterContext * ctx)84 static av_cold void uninit(AVFilterContext *ctx)
85 {
86 RubberBandContext *s = ctx->priv;
87
88 if (s->rbs)
89 rubberband_delete(s->rbs);
90 }
91
query_formats(AVFilterContext * ctx)92 static int query_formats(AVFilterContext *ctx)
93 {
94 AVFilterFormats *formats = NULL;
95 AVFilterChannelLayouts *layouts = NULL;
96 static const enum AVSampleFormat sample_fmts[] = {
97 AV_SAMPLE_FMT_FLTP,
98 AV_SAMPLE_FMT_NONE,
99 };
100 int ret;
101
102 layouts = ff_all_channel_counts();
103 if (!layouts)
104 return AVERROR(ENOMEM);
105 ret = ff_set_common_channel_layouts(ctx, layouts);
106 if (ret < 0)
107 return ret;
108
109 formats = ff_make_format_list(sample_fmts);
110 if (!formats)
111 return AVERROR(ENOMEM);
112 ret = ff_set_common_formats(ctx, formats);
113 if (ret < 0)
114 return ret;
115
116 formats = ff_all_samplerates();
117 if (!formats)
118 return AVERROR(ENOMEM);
119 return ff_set_common_samplerates(ctx, formats);
120 }
121
filter_frame(AVFilterLink * inlink,AVFrame * in)122 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
123 {
124 AVFilterContext *ctx = inlink->dst;
125 RubberBandContext *s = ctx->priv;
126 AVFilterLink *outlink = ctx->outputs[0];
127 AVFrame *out;
128 int ret = 0, nb_samples;
129
130 if (s->first_pts == AV_NOPTS_VALUE)
131 s->first_pts = in->pts;
132
133 rubberband_process(s->rbs, (const float *const *)in->data, in->nb_samples, ff_outlink_get_status(inlink));
134 s->nb_samples_in += in->nb_samples;
135
136 nb_samples = rubberband_available(s->rbs);
137 if (nb_samples > 0) {
138 out = ff_get_audio_buffer(outlink, nb_samples);
139 if (!out) {
140 av_frame_free(&in);
141 return AVERROR(ENOMEM);
142 }
143 out->pts = s->first_pts + av_rescale_q(s->nb_samples_out,
144 (AVRational){ 1, outlink->sample_rate },
145 outlink->time_base);
146 nb_samples = rubberband_retrieve(s->rbs, (float *const *)out->data, nb_samples);
147 out->nb_samples = nb_samples;
148 ret = ff_filter_frame(outlink, out);
149 s->nb_samples_out += nb_samples;
150 }
151
152 av_frame_free(&in);
153 if (ff_inlink_queued_samples(inlink) >= s->nb_samples)
154 ff_filter_set_ready(ctx, 100);
155 return ret < 0 ? ret : nb_samples;
156 }
157
config_input(AVFilterLink * inlink)158 static int config_input(AVFilterLink *inlink)
159 {
160 AVFilterContext *ctx = inlink->dst;
161 RubberBandContext *s = ctx->priv;
162 int opts = s->transients|s->detector|s->phase|s->window|
163 s->smoothing|s->formant|s->opitch|s->channels|
164 RubberBandOptionProcessRealTime;
165
166 if (s->rbs)
167 rubberband_delete(s->rbs);
168 s->rbs = rubberband_new(inlink->sample_rate, inlink->channels, opts, 1. / s->tempo, s->pitch);
169 if (!s->rbs)
170 return AVERROR(ENOMEM);
171
172 s->nb_samples = rubberband_get_samples_required(s->rbs);
173 s->first_pts = AV_NOPTS_VALUE;
174
175 return 0;
176 }
177
activate(AVFilterContext * ctx)178 static int activate(AVFilterContext *ctx)
179 {
180 AVFilterLink *inlink = ctx->inputs[0];
181 AVFilterLink *outlink = ctx->outputs[0];
182 RubberBandContext *s = ctx->priv;
183 AVFrame *in = NULL;
184 int ret;
185
186 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
187
188 ret = ff_inlink_consume_samples(inlink, s->nb_samples, s->nb_samples, &in);
189 if (ret < 0)
190 return ret;
191 if (ret > 0) {
192 ret = filter_frame(inlink, in);
193 if (ret != 0)
194 return ret;
195 }
196
197 FF_FILTER_FORWARD_STATUS(inlink, outlink);
198 FF_FILTER_FORWARD_WANTED(outlink, inlink);
199
200 return FFERROR_NOT_READY;
201 }
202
process_command(AVFilterContext * ctx,const char * cmd,const char * args,char * res,int res_len,int flags)203 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
204 char *res, int res_len, int flags)
205 {
206 RubberBandContext *s = ctx->priv;
207 int ret;
208
209 ret = ff_filter_process_command(ctx, cmd, args, res, res_len, flags);
210 if (ret < 0)
211 return ret;
212
213 rubberband_set_time_ratio(s->rbs, 1. / s->tempo);
214 rubberband_set_pitch_scale(s->rbs, s->pitch);
215 s->nb_samples = rubberband_get_samples_required(s->rbs);
216
217 return 0;
218 }
219
220 static const AVFilterPad rubberband_inputs[] = {
221 {
222 .name = "default",
223 .type = AVMEDIA_TYPE_AUDIO,
224 .config_props = config_input,
225 },
226 { NULL }
227 };
228
229 static const AVFilterPad rubberband_outputs[] = {
230 {
231 .name = "default",
232 .type = AVMEDIA_TYPE_AUDIO,
233 },
234 { NULL }
235 };
236
237 AVFilter ff_af_rubberband = {
238 .name = "rubberband",
239 .description = NULL_IF_CONFIG_SMALL("Apply time-stretching and pitch-shifting."),
240 .query_formats = query_formats,
241 .priv_size = sizeof(RubberBandContext),
242 .priv_class = &rubberband_class,
243 .uninit = uninit,
244 .activate = activate,
245 .inputs = rubberband_inputs,
246 .outputs = rubberband_outputs,
247 .process_command = process_command,
248 };
249