How do I decode cinepak frames

[rvanlaar]

Overview

From a quicktime movie encoded with cinepak I dumped the data per frame to disk.
With a python script I'm able to convert the cinepak frames to PIL images and save those
Script: https://github.com/rvanlaar/QTVR/blob/master/read_cinepak.py

Expected behavior

In the manner of Parsing https://pyav.org/docs/stable/cookbook/basics.html#parsing
I expect to be able to parse packets to frames.

Actual behavior

import av
fh = open("cinepak/nav_13.cinepak", "rb")
codec = av.CodecContext.create("cinepak", "r")
chunk = fh.read()
packets = codec.parse(chunk)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[5], line 1
----> 1 packets = codec.parse(chunk)

File av/codec/context.pyx:343, in av.codec.context.CodecContext.parse()

ValueError: No parser for cinepak

Investigation

I looked at the source code and couldn't see anything which specifically is different.

Research

I have done the following:

• Checked the PyAV documentation
• Searched on Google
• Searched on Stack Overflow
• Looked through old GitHub issues
• Asked on PyAV Gitter
• ... and waited 72 hours for a response.

Additional context

I am working on figuring out QTVR 1 fileformat. For that I would like to be able to pass raw frame data to PyAV to use FFMPEG to decode the frames., as to not have to write a decoder for every different encoding I find.

[rvanlaar]

Dug into the code, it seems to me somethings off with the way the decoder is created...
Tested it with another old codec, namely rpza.

Take the following example

import av

rpza = av.CodecContext.create("rpza", "r")
h264 = av.CodecContext.create("h264", "r")

chunk = b"bogus"

ret = h264.parse(chunk)
assert ret == []

try: 
    rpza.parse(chunk)
except ValueError as err:
    print("Error: couldn't decode")
    raise err

With the output:

Error: couldn't decode
Traceback (most recent call last):
  File "/home/roland/projects/scummvm-director/QTVR/py-parse/spiel.py", line 15, in <module>
    raise err
  File "/home/roland/projects/scummvm-director/QTVR/py-parse/spiel.py", line 12, in <module>
    rpza.parse(chunk)
  File "av/codec/context.pyx", line 343, in av.codec.context.CodecContext.parse
ValueError: No parser for rpza

The exception comes from:
https://github.com/PyAV-Org/PyAV/blob/main/av/codec/context.pyx#L340-L343

https://ffmpeg.org/doxygen/4.4/parser_8c_source.html

[rvanlaar]

Update:
Wrote C code directly interfacing with ffmpeg.
Apparently there is no parser for RPZA.

How to proceed, I still would like to decode single frames.

#include <stdio.h>

#include <libavcodec/avcodec.h>

int main(int argc, char **argv)
{
   const AVCodec *codec;
   AVCodecParserContext *parser;

   codec = avcodec_find_decoder(AV_CODEC_ID_RPZA);
   if (!codec) {
       fprintf(stderr, "Codec not found\n");
       exit(1);
   }
   parser = av_parser_init(codec->id);
   if (!parser) {
       fprintf(stderr, "parser not found\n");
       exit(1);
   }
   printf("codec initialized");
   exit(0);
}

[rvanlaar]

Success.

A little bit of context.
In this quicktime file, I found out the first frame is located at offset 1621 and has a size of 1772

import av

FILENAME = "quicktime.mov"

# constants for this specific movie
# frame 1 starts at:
offset = 1621
# and has a size of
size = 1772
# the frame has these dimensions
image_height = 120
image_width = 120

f = open(FILENAME, "rb")
f.seek(offset)
data = f.read(size)

codec = av.codec.Codec("rpza", "r").create()
codec.width = image_height
codec.height = image_width

packet = av.Packet(data)
frames = codec.decode(packet)
frames[0].to_image().show()

As an idea, create an example for decoding a single frame.