Re: [FFmpeg-devel] [PATCH] Whisper audio filter

[Michael Niedermayer <michael@niedermayer.cc>]

[-- Attachment #1.1: Type: text/plain, Size: 1436 bytes --]

Hi

On Wed, Jul 09, 2025 at 09:23:48AM +0200, Vittorio Palmisano wrote:
> It adds a new audio filter for running audio transcriptions with the whisper model.
> Documentation and examples are included into the patch.
> 
> Signed-off-by: Vittorio Palmisano <vpalmisano@gmail.com>
> ---
>  configure                |   5 +
>  doc/filters.texi         | 101 ++++++++
>  libavfilter/Makefile     |   2 +
>  libavfilter/af_whisper.c | 494 +++++++++++++++++++++++++++++++++++++++
>  libavfilter/allfilters.c |   2 +
>  5 files changed, 604 insertions(+)
>  create mode 100644 libavfilter/af_whisper.c
[...]

> +static void run_transcription(AVFilterContext *ctx, AVDictionary **metadata, int end_pos)
> +{
> +    WhisperContext *wctx = ctx->priv;
> +    end_pos = FFMIN(end_pos, wctx->audio_buffer_fill_size);
> +
> +    if (!wctx->ctx_wsp || end_pos == 0)
> +    {
> +        return;
> +    }
> +
> +    if (!wctx->ctx_wsp)
> +    {
> +        return;
> +    }
> +
> +    float duration = (float)end_pos / WHISPER_SAMPLE_RATE;

In fact float should not be used here
end_pos, audio_buffer_fill_size are all integers
and the timestamp is also integer

exact integer / rational math can and should be used here

thx

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Give a rich man 100$ and he will turn it into 1000$.
Give a poor man 1000$ and he will spend it.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 251 bytes --]

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".