github-mirrors/FFmpeg
1
0
Fork 0
mirror of https://mirror.skon.top/https://github.com/FFmpeg/FFmpeg synced 2026-04-20 21:00:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

avfilter/af_whisper: Add translate parameter

Browse Source
This commit is contained in:
WyattBlue
2026-03-09 00:23:36 -04:00
committed by Gyan Doshi
parent 539fc854e7
commit 482395f830
2 changed files with 20 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
doc/filters.texi
View File

@@ -7743,6 +7743,11 @@ The file path of the downloaded whisper.cpp model (mandatory).
The language to use for transcription ('auto' for auto-detect).
Default value: @code{"auto"}
@item translate
If enabled, translate the transcription from the source language to English. A
multilingual model is required to enable this option.
Default value: @code{"false"}
@item queue
The maximum size that will be queued into the filter before processing the audio
with whisper. Using a small value the audio stream will be processed more often,

15
libavfilter/af_whisper.c
View File

@@ -42,6 +42,7 @@ typedef struct WhisperContext {
const AVClass *class;
char *model_path;
char *language;
bool translate;
bool use_gpu;
int gpu_device;
char *vad_model_path;
@@ -150,6 +151,18 @@ static int init(AVFilterContext *ctx)
wctx->avio_context->direct = AVIO_FLAG_DIRECT;
}
if (!whisper_is_multilingual(wctx->ctx_wsp)) {
if (!wctx->translate && strcmp(wctx->language, "auto") == 0) {
av_log(ctx, AV_LOG_WARNING,
"Multilingual model not provided. Non-English audio may not be correctly transcribed.\n");
} else if (wctx->translate || (strcmp(wctx->language, "auto") != 0 && strcmp(wctx->language, "en") != 0)) {
av_log(ctx, AV_LOG_ERROR,
"%s requested but multilingual model not provided.\n", wctx->translate ? "Translation" : "Transcription");
return AVERROR(ENOSYS);
}
wctx->language = "en";
}
av_log(ctx, AV_LOG_INFO,
"Whisper filter initialized: model: %s lang: %s queue: %" PRId64 " ms\n",
wctx->model_path, wctx->language, wctx->queue / 1000);
@@ -200,6 +213,7 @@ static void run_transcription(AVFilterContext *ctx, AVFrame *frame, int samples)
struct whisper_full_params params = whisper_full_default_params(WHISPER_SAMPLING_GREEDY);
params.language = wctx->language;
params.translate = wctx->translate;
params.n_threads = ff_filter_get_nb_threads(ctx);
params.print_special = 0;
params.print_progress = 0;
@@ -443,6 +457,7 @@ static int query_formats(const AVFilterContext *ctx,
static const AVOption whisper_options[] = {
{ "model", "Path to the whisper.cpp model file", OFFSET(model_path), AV_OPT_TYPE_STRING,.flags = FLAGS },
{ "language", "Language for transcription ('auto' for auto-detect)", OFFSET(language), AV_OPT_TYPE_STRING, {.str = "auto"}, .flags = FLAGS },
{ "translate", "Translate from source language to English", OFFSET(translate), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, .flags = FLAGS },
{ "queue", "Audio queue size", OFFSET(queue), AV_OPT_TYPE_DURATION, {.i64 = 3000000}, 20000, HOURS, .flags = FLAGS },
{ "use_gpu", "Use GPU for processing", OFFSET(use_gpu), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, .flags = FLAGS },
{ "gpu_device", "GPU device to use", OFFSET(gpu_device), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, .flags = FLAGS },