└── ffmpeg_audio.lua


/ffmpeg_audio.lua:
--------------------------------------------------------------------------------
  1 | -- How to decode an audio file with Lua, using LuaJIT FFI and ffmpeg...
  2 | -- Michal Kottman, 2011
  3 | 
  4 | local FILENAME = arg[1] or 'song.mp3'
  5 | local SECTION = print
  6 | 
  7 | SECTION "Initializing the FFI library"
  8 | 
  9 | local ffi = require 'ffi'
 10 | local C = ffi.C
 11 | 
 12 | --[[
 13 | To recreate ffmpeg.h, create a file tmp.h with the following content
 14 | (or more, or less, depending on what you want):
 15 | 
 16 | #include "config.h"
 17 | #include "libavutil/avstring.h"
 18 | #include "libavutil/pixdesc.h"
 19 | #include "libavformat/avformat.h"
 20 | #include "libavdevinputContexte/avdevinputContexte.h"
 21 | #include "libswscale/swscale.h"
 22 | #include "libavcodec/audioconvert.h"
 23 | #include "libavcodec/colorspace.h"
 24 | #include "libavcodec/opt.h"
 25 | #include "libavcodec/avfft.h"
 26 | #include "libavfilter/avfilter.h"
 27 | #include "libavfilter/avfiltergraph.h"
 28 | #include "libavfilter/graphparser.h"
 29 | 
 30 | Then run gcc -E -I $PATH_TO_FFMPEG_SRC tmp.h > ffmpeg.h
 31 | ]]
 32 | 
 33 | local avcodec = ffi.load('avcodec-52')
 34 | local avformat = ffi.load('avformat-52')
 35 | local avutil = ffi.load('avutil-50')
 36 | local header = assert(io.open('ffmpeg.h')):read('*a')
 37 | ffi.cdef(header)
 38 | 
 39 | function avAssert(err)
 40 | 	if err < 0 then
 41 | 		local errbuf = ffi.new("uint8_t[256]")
 42 | 		local ret = avutil.av_strerror(err, errbuf, 256)
 43 | 		if ret ~= -1 then
 44 | 			error(ffi.string(errbuf), 2)
 45 | 		else
 46 | 			error('Unknown AV error: '..tostring(ret), 2)
 47 | 		end
 48 | 	end
 49 | 	return err
 50 | end
 51 | 
 52 | SECTION "Initializing the avcodec and avformat libraries"
 53 | 
 54 | avcodec.avcodec_init()
 55 | avcodec.avcodec_register_all()
 56 | avformat.av_register_all()
 57 | 
 58 | SECTION "Opening file"
 59 | 
 60 | local pinputContext = ffi.new("AVFormatContext*[1]")
 61 | avAssert(avformat.av_open_input_file(pinputContext, FILENAME, nil, 0, nil))
 62 | local inputContext = pinputContext[0]
 63 | 
 64 | avAssert(avformat.av_find_stream_info(inputContext))
 65 | 
 66 | SECTION "Finding audio stream"
 67 | 
 68 | local audioCtx
 69 | local nStreams = tonumber(inputContext.nb_streams)
 70 | for i=1,nStreams do
 71 | 	local stream = inputContext.streams[i-1]
 72 | 	local ctx = stream.codec
 73 | 	if ctx.codec_type == C.AVMEDIA_TYPE_AUDIO then
 74 | 		local codec = avcodec.avcodec_find_decoder(ctx.codec_id)
 75 | 		avAssert(avcodec.avcodec_open(ctx, codec))
 76 | 		audioCtx = ctx
 77 | 	end
 78 | end
 79 | if not audioCtx then error('Unable to find audio stream') end
 80 | 
 81 | print("Bitrate:", tonumber(audioCtx.bit_rate))
 82 | print("Channels:", tonumber(audioCtx.channels))
 83 | print("Sample rate:", tonumber(audioCtx.sample_rate))
 84 | print("Sample type:", ({[0]="u8", "s16", "s32", "flt", "dbl"})[audioCtx.sample_fmt])
 85 | 
 86 | SECTION "Decoding"
 87 | 
 88 | local AVCODEC_MAX_AUDIO_FRAME_SIZE = 192000
 89 | 
 90 | local packet = ffi.new("AVPacket")
 91 | local temp_frame = ffi.new("int16_t[?]", AVCODEC_MAX_AUDIO_FRAME_SIZE)
 92 | local frame_size = ffi.new("int[1]")
 93 | 
 94 | local all_samples = {}
 95 | local total_samples = 0
 96 | 
 97 | while tonumber(avformat.url_feof(inputContext.pb)) == 0 do
 98 | 	local ret = avAssert(avformat.av_read_frame(inputContext, packet))
 99 | 
100 | 	frame_size[0] = AVCODEC_MAX_AUDIO_FRAME_SIZE
101 | 	local n = avcodec.avcodec_decode_audio3(audioCtx, temp_frame, frame_size, packet)
102 | 	if n == -1 then break
103 | 	elseif n < 0 then avAssert(n) end
104 | 
105 | 	local size = tonumber(frame_size[0])/2 -- frame_size is in bytes
106 | 	local frame = ffi.new("int16_t[?]", size)
107 | 	ffi.copy(frame, temp_frame, size*2)
108 | 	all_samples[#all_samples + 1] = frame
109 | 	total_samples = total_samples + size
110 | end
111 | 
112 | SECTION "Merging samples"
113 | 
114 | local samples = ffi.new("int16_t[?]", total_samples)
115 | local offset = 0
116 | for _,s in ipairs(all_samples) do
117 | 	local size = ffi.sizeof(s)
118 | 	ffi.copy(samples + offset, s, size)
119 | 	offset = offset + size/2
120 | end
121 | 
122 | SECTION "Processing"
123 | 
124 | -- The `samples` array is now ready for some processing! :)
125 | 
126 | -- ... like writing it raw to a file
127 | 
128 | local out = assert(io.open('samples.raw', 'wb'))
129 | local size = ffi.sizeof(samples)
130 | out:write(ffi.string(samples, size))
131 | out:close()
132 | 
133 | -- Now you can open it in any audio processing program to see that it works.
134 | -- In Audacity: Project -> Import Raw Data (and fill out according to info)
135 | 


--------------------------------------------------------------------------------