[FFmpeg-devel] [PATCH v2 2/3] avcodec: estimate output bitrate for ffv1/huffyuv codecs

[Tobias Rapp]

On 03.03.2017 16:14, Michael Niedermayer wrote:
> On Mon, Feb 06, 2017 at 01:33:19PM +0100, Tobias Rapp wrote:
>> Allows to get a more realistic total bitrate (and estimated file size)
>> in avi_write_header. Previously a static default value of 200k was
>> assumed.
>>
>> Signed-off-by: Tobias Rapp <t.rapp at noa-archive.com>
>> ---
>>  [...]
>> diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c
>> index 35f54c6..7138cc0 100644
>> --- a/libavcodec/ffv1enc.c
>> +++ b/libavcodec/ffv1enc.c
>> @@ -894,6 +894,10 @@ slices_ok:
>>              }
>>      }
>>
>> +    // estimate bitrate assuming 40% compression
>> +    avctx->bit_rate = ff_guess_coded_bitrate(avctx) * 2 / 5;
>> +    avctx->bit_rate_tolerance = avctx->bit_rate;
>> +
>>      return 0;
>>  }
>>
>> diff --git a/libavcodec/huffyuvenc.c b/libavcodec/huffyuvenc.c
>> index 89639b7..8911408 100644
>> --- a/libavcodec/huffyuvenc.c
>> +++ b/libavcodec/huffyuvenc.c
>> @@ -446,6 +446,10 @@ FF_ENABLE_DEPRECATION_WARNINGS
>>
>>      s->picture_number=0;
>>
>> +    // estimate bitrate assuming 60% compression
>> +    avctx->bit_rate = ff_guess_coded_bitrate(avctx) * 3 / 5;
>> +    avctx->bit_rate_tolerance = avctx->bit_rate;
>
> these violate the API
>
> See avcodec.h:
>     /**
>      * number of bits the bitstream is allowed to diverge from the reference.
>      *           the reference can be CBR (for CBR pass1) or VBR (for pass2)
>      * - encoding: Set by user; unused for constant quantizer encoding.
>      * - decoding: unused
>      */
>     int bit_rate_tolerance;
>
> also the compression rate depends on the material these numbers are
> basically random

I agree that the bitrate can vary much depending on the type of video 
material (animation, film scanning, noise, etc) and my intent was to 
signal that with +/- bit_rate_tolerance. Anyway the 40%/60% compression 
numbers roughly match my experience with analogue ingest.

Will post version 3 of the patch-set without updates to bit_rate_tolerance.

Regards,
Tobias