Merged in revid-audio (pull request #204)

Revid audio

Approved-by: kortschak <dan@kortschak.io>

cmd/revid-cli/main.go

@@ -8,6 +8,7 @@ DESCRIPTION
 AUTHORS
   Saxon A. Nelson-Milton <saxon@ausocean.org>
   Jack Richardson <jack@ausocean.org>
+  Trek Hopton <trek@ausocean.org>
 
 LICENSE
   revid-cli is Copyright (C) 2017-2018 the Australian Ocean Lab (AusOcean)
@@ -26,6 +27,7 @@ LICENSE
   along with revid in gpl.txt. If not, see http://www.gnu.org/licenses.
 */
 
+// revid-cli is a command line interface for revid.
 package main
 
 import (
@@ -36,6 +38,7 @@ import (
 	"strings"
 	"time"
 
+	"bitbucket.org/ausocean/av/codec/codecutil"
 	"bitbucket.org/ausocean/av/container/mts"
 	"bitbucket.org/ausocean/av/container/mts/meta"
 	"bitbucket.org/ausocean/av/revid"
@@ -105,9 +108,9 @@ func handleFlags() revid.Config {
 	var (
 		cpuprofile = flag.String("cpuprofile", "", "write cpu profile to `file`")
 
-		inputPtr              = flag.String("Input", "", "The input type: Raspivid, File, Webcam, RTSP")
+		inputCodecPtr         = flag.String("InputCodec", "", "The codec of the input: H264, Mjpeg, PCM, ADPCM")
+		inputPtr              = flag.String("Input", "", "The input type: Raspivid, File, v4l, Audio, RTSP")
 		rtspURLPtr            = flag.String("RTSPURL", "", "The URL for an RTSP server.")
-		inputCodecPtr         = flag.String("InputCodec", "", "The codec of the input: H264, Mjpeg")
 		quantizePtr           = flag.Bool("Quantize", false, "Quantize input (non-variable bitrate)")
 		verbosityPtr          = flag.String("Verbosity", "Info", "Verbosity: Debug, Info, Warning, Error, Fatal")
 		rtpAddrPtr            = flag.String("RtpAddr", "", "Rtp destination address: <IP>:<port> (port is generally 6970-6999)")
@@ -131,6 +134,12 @@ func handleFlags() revid.Config {
 		saturationPtr         = flag.Int("Saturation", 0, "Set Saturation. (100-100)")
 		exposurePtr           = flag.String("Exposure", "auto", "Set exposure mode. ("+strings.Join(revid.ExposureModes[:], ",")+")")
 		autoWhiteBalancePtr   = flag.String("Awb", "auto", "Set automatic white balance mode. ("+strings.Join(revid.AutoWhiteBalanceModes[:], ",")+")")
+
+		// Audio specific flags.
+		sampleRatePtr = flag.Int("SampleRate", 48000, "Sample rate of recorded audio")
+		channelsPtr   = flag.Int("Channels", 1, "Record in Mono or Stereo (1 or 2)")
+		recPeriodPtr  = flag.Float64("recPeriod", 1, "How many seconds to record at a time")
+		bitDepthPtr   = flag.Int("bitDepth", 16, "Bit Depth to record audio at.")
 	)
 
 	var outputs flagStrings
@@ -179,6 +188,8 @@ func handleFlags() revid.Config {
 		cfg.Input = revid.V4L
 	case "File":
 		cfg.Input = revid.File
+	case "Audio":
+		cfg.Input = revid.Audio
 	case "RTSP":
 		cfg.Input = revid.RTSP
 	case "":
@@ -188,12 +199,23 @@ func handleFlags() revid.Config {
 
 	switch *inputCodecPtr {
 	case "H264":
-		cfg.InputCodec = revid.H264
+		cfg.InputCodec = codecutil.H264
+	case "PCM":
+		cfg.InputCodec = codecutil.PCM
+	case "ADPCM":
+		cfg.InputCodec = codecutil.ADPCM
 	case "":
 	default:
 		log.Log(logger.Error, pkg+"bad input codec argument")
 	}
 
+	switch *inputPtr {
+	case "Audio":
+		cfg.WriteRate = 1.0 / (*recPeriodPtr)
+	default:
+		cfg.WriteRate = float64(*frameRatePtr)
+	}
+
 	for _, o := range outputs {
 		switch o {
 		case "File":
@@ -235,6 +257,10 @@ func handleFlags() revid.Config {
 	cfg.Saturation = *saturationPtr
 	cfg.Exposure = *exposurePtr
 	cfg.AutoWhiteBalance = *autoWhiteBalancePtr
+	cfg.SampleRate = *sampleRatePtr
+	cfg.Channels = *channelsPtr
+	cfg.RecPeriod = *recPeriodPtr
+	cfg.BitDepth = *bitDepthPtr
 
 	return cfg
 }

codec/codecutil/lex.go

@@ -0,0 +1,84 @@
+/*
+NAME
+  lex.go
+
+AUTHOR
+  Trek Hopton <trek@ausocean.org>
+
+LICENSE
+  This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
+
+  It is free software: you can redistribute it and/or modify them
+  under the terms of the GNU General Public License as published by the
+  Free Software Foundation, either version 3 of the License, or (at your
+  option) any later version.
+
+  It is distributed in the hope that it will be useful, but WITHOUT
+  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+  for more details.
+
+  You should have received a copy of the GNU General Public License in gpl.txt.
+  If not, see [GNU licenses](http://www.gnu.org/licenses).
+*/
+
+package codecutil
+
+import (
+	"fmt"
+	"io"
+	"time"
+)
+
+// ByteLexer is used to lex bytes using a buffer size which is configured upon construction.
+type ByteLexer struct {
+	bufSize *int
+}
+
+// NewByteLexer returns a pointer to a ByteLexer with the given buffer size.
+func NewByteLexer(bufSize *int) *ByteLexer {
+	return &ByteLexer{bufSize: bufSize}
+}
+
+// zeroTicks can be used to create an instant ticker.
+var zeroTicks chan time.Time
+
+func init() {
+	zeroTicks = make(chan time.Time)
+	close(zeroTicks)
+}
+
+// Lex reads *l.bufSize bytes from src and writes them to dst every d seconds.
+func (l *ByteLexer) Lex(dst io.Writer, src io.Reader, d time.Duration) error {
+	if l.bufSize == nil {
+		return fmt.Errorf("buffer size has not been set")
+	}
+	bufSize := *l.bufSize
+	if bufSize <= 0 {
+		return fmt.Errorf("invalid buffer size: %v", bufSize)
+	}
+	if d < 0 {
+		return fmt.Errorf("invalid delay: %v", d)
+	}
+
+	var ticker *time.Ticker
+	if d == 0 {
+		ticker = &time.Ticker{C: zeroTicks}
+	} else {
+		ticker = time.NewTicker(d)
+		defer ticker.Stop()
+	}
+
+	buf := make([]byte, bufSize)
+	for {
+		<-ticker.C
+		off, err := src.Read(buf)
+		if err != nil {
+			return err
+		}
+		_, err = dst.Write(buf[:off])
+		if err != nil {
+			return err
+		}
+	}
+}

codec/codecutil/lex_test.go

@@ -0,0 +1,65 @@
+/*
+NAME
+  lex_test.go
+
+AUTHOR
+  Trek Hopton <trek@ausocean.org>
+
+LICENSE
+  This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
+
+  It is free software: you can redistribute it and/or modify them
+  under the terms of the GNU General Public License as published by the
+  Free Software Foundation, either version 3 of the License, or (at your
+  option) any later version.
+
+  It is distributed in the hope that it will be useful, but WITHOUT
+  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+  for more details.
+
+  You should have received a copy of the GNU General Public License in gpl.txt.
+  If not, see [GNU licenses](http://www.gnu.org/licenses).
+*/
+
+package codecutil
+
+import (
+	"bytes"
+	"io"
+	"strconv"
+	"testing"
+	"time"
+)
+
+var lexTests = []struct {
+	data    []byte
+	t       time.Duration
+	n       int
+	isValid bool // Whether or not this test should fail.
+}{
+	{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, 4, true},
+	{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, 3, true},
+	{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, 0, 2, true},
+	{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, 0, 1, true},
+	{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Nanosecond, 0, false},
+	{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, -1, false},
+	{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, 15, true},
+}
+
+func TestByteLexer(t *testing.T) {
+	for i, tt := range lexTests {
+		t.Run(strconv.Itoa(i), func(t *testing.T) {
+			dst := bytes.NewBuffer([]byte{})
+			l := NewByteLexer(&tt.n)
+			err := l.Lex(dst, bytes.NewReader(tt.data), tt.t)
+			if err != nil && err != io.EOF {
+				if tt.isValid {
+					t.Errorf("unexpected error: %v", err)
+				}
+			} else if !bytes.Equal(dst.Bytes(), tt.data) {
+				t.Errorf("data before and after lex are not equal: want %v, got %v", tt.data, dst.Bytes())
+			}
+		})
+	}
+}

codec/codecutil/list.go

@@ -0,0 +1,43 @@
+/*
+NAME
+  list.go
+
+AUTHOR
+  Trek Hopton <trek@ausocean.org>
+
+LICENSE
+  This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
+
+  It is free software: you can redistribute it and/or modify them
+  under the terms of the GNU General Public License as published by the
+  Free Software Foundation, either version 3 of the License, or (at your
+  option) any later version.
+
+  It is distributed in the hope that it will be useful, but WITHOUT
+  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+  for more details.
+
+  You should have received a copy of the GNU General Public License in gpl.txt.
+  If not, see [GNU licenses](http://www.gnu.org/licenses).
+*/
+
+package codecutil
+
+// numCodecs is the number of entries in the list of codecs.
+const numCodecs = 5
+
+// A global list containing all available codecs for reference in any application.
+// When adding or removing a codec from this list, the numCodecs const must be updated.
+const (
+	PCM = iota
+	ADPCM
+	H264
+	H265
+	MJPEG
+)
+
+// IsValid recieves an int representing a codec and checks if it is valid.
+func IsValid(codec uint8) bool {
+	return 0 <= codec && codec < numCodecs
+}

codec/pcm/pcm.go

@@ -35,20 +35,20 @@ import (
 	"github.com/yobert/alsa"
 )
 
-// Resample takes an alsa.Buffer (b) and resamples the pcm audio data to 'rate' Hz and returns the resulting pcm.
+// Resample takes alsa.Buffer b and resamples the pcm audio data to 'rate' Hz and returns an alsa.Buffer with the resampled data.
-// If an error occurs, an error will be returned along with the original b's data.
 // Notes:
 // 	- Currently only downsampling is implemented and b's rate must be divisible by 'rate' or an error will occur.
 // 	- If the number of bytes in b.Data is not divisible by the decimation factor (ratioFrom), the remaining bytes will
 // 	  not be included in the result. Eg. input of length 480002 downsampling 6:1 will result in output length 80000.
-func Resample(b alsa.Buffer, rate int) ([]byte, error) {
+func Resample(b alsa.Buffer, rate int) (alsa.Buffer, error) {
-	fromRate := b.Format.Rate
+	if b.Format.Rate == rate {
-	if fromRate == rate {
+		return b, nil
-		return b.Data, nil
+	}
-	} else if fromRate < 0 {
+	if b.Format.Rate < 0 {
-		return nil, fmt.Errorf("Unable to convert from: %v Hz", fromRate)
+		return alsa.Buffer{}, fmt.Errorf("Unable to convert from: %v Hz", b.Format.Rate)
-	} else if rate < 0 {
+	}
-		return nil, fmt.Errorf("Unable to convert to: %v Hz", rate)
+	if rate < 0 {
+		return alsa.Buffer{}, fmt.Errorf("Unable to convert to: %v Hz", rate)
 	}
 
 	// The number of bytes in a sample.
@@ -59,22 +59,22 @@ func Resample(b alsa.Buffer, rate int) ([]byte, error) {
 	case alsa.S16_LE:
 		sampleLen = 2 * b.Format.Channels
 	default:
-		return nil, fmt.Errorf("Unhandled ALSA format: %v", b.Format.SampleFormat)
+		return alsa.Buffer{}, fmt.Errorf("Unhandled ALSA format: %v", b.Format.SampleFormat)
 	}
 	inPcmLen := len(b.Data)
 
 	// Calculate sample rate ratio ratioFrom:ratioTo.
-	rateGcd := gcd(rate, fromRate)
+	rateGcd := gcd(rate, b.Format.Rate)
-	ratioFrom := fromRate / rateGcd
+	ratioFrom := b.Format.Rate / rateGcd
 	ratioTo := rate / rateGcd
 
 	// ratioTo = 1 is the only number that will result in an even sampling.
 	if ratioTo != 1 {
-		return nil, fmt.Errorf("unhandled from:to rate ratio %v:%v: 'to' must be 1", ratioFrom, ratioTo)
+		return alsa.Buffer{}, fmt.Errorf("unhandled from:to rate ratio %v:%v: 'to' must be 1", ratioFrom, ratioTo)
 	}
 
 	newLen := inPcmLen / ratioFrom
-	result := make([]byte, 0, newLen)
+	resampled := make([]byte, 0, newLen)
 
 	// For each new sample to be generated, loop through the respective 'ratioFrom' samples in 'b.Data' to add them
 	// up and average them. The result is the new sample.
@@ -96,19 +96,28 @@ func Resample(b alsa.Buffer, rate int) ([]byte, error) {
 		case alsa.S16_LE:
 			binary.LittleEndian.PutUint16(bAvg, uint16(avg))
 		}
-		result = append(result, bAvg...)
+		resampled = append(resampled, bAvg...)
 	}
-	return result, nil
+
+	// Return a new alsa.Buffer with resampled data.
+	return alsa.Buffer{
+		Format: alsa.BufferFormat{
+			Channels:     b.Format.Channels,
+			SampleFormat: b.Format.SampleFormat,
+			Rate:         rate,
+		},
+		Data: resampled,
+	}, nil
 }
 
 // StereoToMono returns raw mono audio data generated from only the left channel from
 // the given stereo recording (ALSA buffer)
-// if an error occurs, an error will be returned along with the original stereo data.
+func StereoToMono(b alsa.Buffer) (alsa.Buffer, error) {
-func StereoToMono(b alsa.Buffer) ([]byte, error) {
 	if b.Format.Channels == 1 {
-		return b.Data, nil
+		return b, nil
-	} else if b.Format.Channels != 2 {
+	}
-		return nil, fmt.Errorf("Audio is not stereo or mono, it has %v channels", b.Format.Channels)
+	if b.Format.Channels != 2 {
+		return alsa.Buffer{}, fmt.Errorf("Audio is not stereo or mono, it has %v channels", b.Format.Channels)
 	}
 
 	var stereoSampleBytes int
@@ -118,7 +127,7 @@ func StereoToMono(b alsa.Buffer) ([]byte, error) {
 	case alsa.S16_LE:
 		stereoSampleBytes = 4
 	default:
-		return nil, fmt.Errorf("Unhandled ALSA format %v", b.Format.SampleFormat)
+		return alsa.Buffer{}, fmt.Errorf("Unhandled ALSA format %v", b.Format.SampleFormat)
 	}
 
 	recLength := len(b.Data)
@@ -134,7 +143,15 @@ func StereoToMono(b alsa.Buffer) ([]byte, error) {
 		}
 	}
 
-	return mono, nil
+	// Return a new alsa.Buffer with resampled data.
+	return alsa.Buffer{
+		Format: alsa.BufferFormat{
+			Channels:     1,
+			SampleFormat: b.Format.SampleFormat,
+			Rate:         b.Format.Rate,
+		},
+		Data: mono,
+	}, nil
 }
 
 // gcd is used for calculating the greatest common divisor of two positive integers, a and b.

codec/pcm/pcm_test.go

@@ -71,7 +71,7 @@ func TestResample(t *testing.T) {
 	}
 
 	// Compare result with expected.
-	if !bytes.Equal(resampled, exp) {
+	if !bytes.Equal(resampled.Data, exp) {
 		t.Error("Resampled data does not match expected result.")
 	}
 }
@@ -112,7 +112,7 @@ func TestStereoToMono(t *testing.T) {
 	}
 
 	// Compare result with expected.
-	if !bytes.Equal(mono, exp) {
+	if !bytes.Equal(mono.Data, exp) {
 		t.Error("Converted data does not match expected result.")
 	}
 }

container/mts/encoder.go

@@ -89,11 +89,6 @@ var (
 )
 
 const (
-	sdtPid        = 17
-	patPid        = 0
-	pmtPid        = 4096
-	videoPid      = 256
-	audioPid      = 210
 	H264ID        = 27
 	H265ID        = 36
 	audioStreamID = 0xc0 // First audio stream ID.
@@ -151,13 +146,13 @@ func NewEncoder(dst io.WriteCloser, rate float64, mediaType int) *Encoder {
 	var sid byte
 	switch mediaType {
 	case EncodeAudio:
-		mPid = audioPid
+		mPid = AudioPid
 		sid = audioStreamID
 	case EncodeH265:
-		mPid = videoPid
+		mPid = VideoPid
 		sid = H265ID
 	case EncodeH264:
-		mPid = videoPid
+		mPid = VideoPid
 		sid = H264ID
 	}
 
@@ -187,8 +182,8 @@ func NewEncoder(dst io.WriteCloser, rate float64, mediaType int) *Encoder {
 		streamID: sid,
 
 		continuity: map[int]byte{
-			patPid: 0,
+			PatPid: 0,
-			pmtPid: 0,
+			PmtPid: 0,
 			mPid:   0,
 		},
 	}
@@ -214,7 +209,7 @@ func (e *Encoder) TimeBasedPsi(b bool, sendCount int) {
 	e.pktCount = e.psiSendCount
 }
 
-// Write implements io.Writer. Write takes raw h264 and encodes into MPEG-TS,
+// Write implements io.Writer. Write takes raw video or audio data and encodes into MPEG-TS,
 // then sending it to the encoder's io.Writer destination.
 func (e *Encoder) Write(data []byte) (int, error) {
 	now := time.Now()

container/mts/encoder_test.go

@@ -199,7 +199,7 @@ func TestEncodePcm(t *testing.T) {
 	for i+PacketSize <= len(clip) {
 
 		// Check MTS packet
-		if !(pkt.PID() == audioPid) {
+		if pkt.PID() != AudioPid {
 			i += PacketSize
 			if i+PacketSize <= len(clip) {
 				copy(pkt[:], clip[i:i+PacketSize])

container/mts/mpegts.go

@@ -47,6 +47,7 @@ const (
 	PatPid   = 0
 	PmtPid   = 4096
 	VideoPid = 256
+	AudioPid = 210
 )
 
 // StreamID is the id of the first stream.

container/mts/mpegts_test.go

@@ -82,7 +82,7 @@ func TestGetPTSRange1(t *testing.T) {
 		curTime += interval
 	}
 
-	got, err := GetPTSRange(clip.Bytes(), videoPid)
+	got, err := GetPTSRange(clip.Bytes(), VideoPid)
 	if err != nil {
 		t.Fatalf("did not expect error getting PTS range: %v", err)
 	}
@@ -142,7 +142,7 @@ func writeFrame(b *bytes.Buffer, frame []byte, pts uint64) error {
 	for len(buf) != 0 {
 		pkt := Packet{
 			PUSI: pusi,
-			PID:  videoPid,
+			PID:  VideoPid,
 			RAI:  pusi,
 			CC:   0,
 			AFC:  hasAdaptationField | hasPayload,

exp/pcm/resample/resample.go

@@ -81,9 +81,9 @@ func main() {
 	}
 
 	// Save resampled to file.
-	err = ioutil.WriteFile(outPath, resampled, 0644)
+	err = ioutil.WriteFile(outPath, resampled.Data, 0644)
 	if err != nil {
 		log.Fatal(err)
 	}
-	fmt.Println("Encoded and wrote", len(resampled), "bytes to file", outPath)
+	fmt.Println("Encoded and wrote", len(resampled.Data), "bytes to file", outPath)
 }

exp/pcm/stereo-to-mono/stereo-to-mono.go

@@ -77,9 +77,9 @@ func main() {
 	}
 
 	// Save mono to file.
-	err = ioutil.WriteFile(outPath, mono, 0644)
+	err = ioutil.WriteFile(outPath, mono.Data, 0644)
 	if err != nil {
 		log.Fatal(err)
 	}
-	fmt.Println("Encoded and wrote", len(mono), "bytes to file", outPath)
+	fmt.Println("Encoded and wrote", len(mono.Data), "bytes to file", outPath)
 }

input/audio/audio.go

@@ -0,0 +1,464 @@
+/*
+NAME
+  audio.go
+
+AUTHOR
+  Alan Noble <alan@ausocean.org>
+  Trek Hopton <trek@ausocean.org>
+
+LICENSE
+  This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
+
+  It is free software: you can redistribute it and/or modify them
+  under the terms of the GNU General Public License as published by the
+  Free Software Foundation, either version 3 of the License, or (at your
+  option) any later version.
+
+  It is distributed in the hope that it will be useful, but WITHOUT
+  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+  for more details.
+
+  You should have received a copy of the GNU General Public License in gpl.txt.
+  If not, see [GNU licenses](http://www.gnu.org/licenses).
+*/
+
+// Package audio provides access to input from audio devices.
+package audio
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"sync"
+	"time"
+
+	"github.com/yobert/alsa"
+
+	"bitbucket.org/ausocean/av/codec/adpcm"
+	"bitbucket.org/ausocean/av/codec/codecutil"
+	"bitbucket.org/ausocean/av/codec/pcm"
+	"bitbucket.org/ausocean/utils/logger"
+	"bitbucket.org/ausocean/utils/ring"
+)
+
+const (
+	pkg               = "audio: "
+	rbTimeout         = 100 * time.Millisecond
+	rbNextTimeout     = 100 * time.Millisecond
+	rbLen             = 200
+	defaultSampleRate = 48000
+)
+
+// "running" means the input goroutine is reading from the ALSA device and writing to the ringbuffer.
+// "paused" means the input routine is sleeping until unpaused or stopped.
+// "stopped" means the input routine is stopped and the ALSA device is closed.
+const (
+	running = iota + 1
+	paused
+	stopped
+)
+
+// Device holds everything we need to know about the audio input stream and implements io.Reader.
+type Device struct {
+	l         Logger       // Logger for device's routines to log to.
+	mode      uint8        // Operating mode, either running, paused, or stopped.
+	mu        sync.Mutex   // Provides synchronisation when changing modes concurrently.
+	title     string       // Name of audio title, or empty for the default title.
+	dev       *alsa.Device // ALSA's Audio input device.
+	ab        alsa.Buffer  // ALSA's buffer.
+	rb        *ring.Buffer // Our buffer.
+	chunkSize int          // This is the number of bytes that will be stored in rb at a time.
+	*Config                // Configuration parameters for this device.
+}
+
+// Config provides parameters used by Device.
+type Config struct {
+	SampleRate int
+	Channels   int
+	BitDepth   int
+	RecPeriod  float64
+	Codec      uint8
+}
+
+// Logger enables any implementation of a logger to be used.
+// TODO: Make this part of the logger package.
+type Logger interface {
+	SetLevel(int8)
+	Log(level int8, message string, params ...interface{})
+}
+
+// OpenError is used to determine whether an error has originated from attempting to open a device.
+type OpenError error
+
+// NewDevice initializes and returns an Device which can be started, read from, and stopped.
+func NewDevice(cfg *Config, l Logger) (*Device, error) {
+
+	err := validate(cfg)
+	if err != nil {
+		return nil, err
+	}
+
+	d := &Device{
+		Config: cfg,
+		l:      l,
+	}
+
+	// Open the requested audio device.
+	err = d.open()
+	if err != nil {
+		d.l.Log(logger.Error, pkg+"failed to open device")
+		return nil, err
+	}
+
+	// Setup the device to record with desired period.
+	d.ab = d.dev.NewBufferDuration(time.Duration(d.RecPeriod * float64(time.Second)))
+
+	// Account for channel conversion.
+	chunkSize := float64(len(d.ab.Data) / d.dev.BufferFormat().Channels * d.Channels)
+
+	// Account for resampling.
+	chunkSize = (chunkSize / float64(d.dev.BufferFormat().Rate)) * float64(d.SampleRate)
+	if chunkSize < 1 {
+		return nil, errors.New("given Config parameters are too small")
+	}
+
+	// Account for codec conversion.
+	if d.Codec == codecutil.ADPCM {
+		d.chunkSize = adpcm.EncBytes(int(chunkSize))
+	} else {
+		d.chunkSize = int(chunkSize)
+	}
+
+	// Create ring buffer with appropriate chunk size.
+	d.rb = ring.NewBuffer(rbLen, d.chunkSize, rbTimeout)
+
+	// Start device in paused mode.
+	d.mode = paused
+	go d.input()
+
+	return d, nil
+}
+
+// Start will start recording audio and writing to the ringbuffer.
+// Once a Device has been stopped it cannot be started again. This is likely to change in future.
+func (d *Device) Start() error {
+	d.mu.Lock()
+	mode := d.mode
+	d.mu.Unlock()
+	switch mode {
+	case paused:
+		d.mu.Lock()
+		d.mode = running
+		d.mu.Unlock()
+		return nil
+	case stopped:
+		// TODO(Trek): Make this reopen device and start recording.
+		return errors.New("device is stopped")
+	case running:
+		return nil
+	default:
+		return fmt.Errorf("invalid mode: %d", mode)
+	}
+}
+
+// Stop will stop recording audio and close the device.
+// Once a Device has been stopped it cannot be started again. This is likely to change in future.
+func (d *Device) Stop() {
+	d.mu.Lock()
+	d.mode = stopped
+	d.mu.Unlock()
+}
+
+// ChunkSize returns the number of bytes written to the ringbuffer per d.RecPeriod.
+func (d *Device) ChunkSize() int {
+	return d.chunkSize
+}
+
+// validate checks if Config parameters are valid and returns an error if they are not.
+func validate(c *Config) error {
+	if c.SampleRate <= 0 {
+		return fmt.Errorf("invalid sample rate: %v", c.SampleRate)
+	}
+	if c.Channels <= 0 {
+		return fmt.Errorf("invalid number of channels: %v", c.Channels)
+	}
+	if c.BitDepth <= 0 {
+		return fmt.Errorf("invalid bitdepth: %v", c.BitDepth)
+	}
+	if c.RecPeriod <= 0 {
+		return fmt.Errorf("invalid recording period: %v", c.RecPeriod)
+	}
+	if !codecutil.IsValid(c.Codec) {
+		return errors.New("invalid codec")
+	}
+	return nil
+}
+
+// open the recording device with the given name and prepare it to record.
+// If name is empty, the first recording device is used.
+func (d *Device) open() error {
+	// Close any existing device.
+	if d.dev != nil {
+		d.l.Log(logger.Debug, pkg+"closing device", "title", d.title)
+		d.dev.Close()
+		d.dev = nil
+	}
+
+	// Open sound card and open recording device.
+	d.l.Log(logger.Debug, pkg+"opening sound card")
+	cards, err := alsa.OpenCards()
+	if err != nil {
+		return OpenError(err)
+	}
+	defer alsa.CloseCards(cards)
+
+	d.l.Log(logger.Debug, pkg+"finding audio device")
+	for _, card := range cards {
+		devices, err := card.Devices()
+		if err != nil {
+			continue
+		}
+		for _, dev := range devices {
+			if dev.Type != alsa.PCM || !dev.Record {
+				continue
+			}
+			if dev.Title == d.title || d.title == "" {
+				d.dev = dev
+				break
+			}
+		}
+	}
+	if d.dev == nil {
+		return OpenError(errors.New("no audio device found"))
+	}
+
+	d.l.Log(logger.Debug, pkg+"opening audio device", "title", d.dev.Title)
+	err = d.dev.Open()
+	if err != nil {
+		return OpenError(err)
+	}
+
+	// 2 channels is what most devices need to record in. If mono is requested,
+	// the recording will be converted in formatBuffer().
+	channels, err := d.dev.NegotiateChannels(2)
+	if err != nil {
+		return OpenError(err)
+	}
+	d.l.Log(logger.Debug, pkg+"alsa device channels set", "channels", channels)
+
+	// Try to negotiate a rate to record in that is divisible by the wanted rate
+	// so that it can be easily downsampled to the wanted rate.
+	// rates is a slice of common sample rates including the standard for CD (44100Hz) and standard for professional audio recording (48000Hz).
+	// Note: if a card thinks it can record at a rate but can't actually, this can cause a failure.
+	// Eg. the audioinjector sound card is supposed to record at 8000Hz and 16000Hz but it can't due to a firmware issue,
+	// a fix for this is to remove 8000 and 16000 from the rates slice.
+	var rates = [8]int{8000, 16000, 32000, 44100, 48000, 88200, 96000, 192000}
+
+	var rate int
+	foundRate := false
+	for r := range rates {
+		if r < d.SampleRate {
+			continue
+		}
+		if r%d.SampleRate == 0 {
+			rate, err = d.dev.NegotiateRate(r)
+			if err == nil {
+				foundRate = true
+				d.l.Log(logger.Debug, pkg+"alsa device sample rate set", "rate", rate)
+				break
+			}
+		}
+	}
+
+	// If no easily divisible rate is found, then use the default rate.
+	if !foundRate {
+		d.l.Log(logger.Warning, pkg+"Unable to sample at requested rate, default used.", "rateRequested", d.SampleRate)
+		rate, err = d.dev.NegotiateRate(defaultSampleRate)
+		if err != nil {
+			return OpenError(err)
+		}
+		d.l.Log(logger.Debug, pkg+"alsa device sample rate set", "rate", rate)
+	}
+
+	var aFmt alsa.FormatType
+	switch d.BitDepth {
+	case 16:
+		aFmt = alsa.S16_LE
+	case 32:
+		aFmt = alsa.S32_LE
+	default:
+		return OpenError(fmt.Errorf("unsupported sample bits %v", d.BitDepth))
+	}
+	devFmt, err := d.dev.NegotiateFormat(aFmt)
+	if err != nil {
+		return err
+	}
+	var bitdepth int
+	switch devFmt {
+	case alsa.S16_LE:
+		bitdepth = 16
+	case alsa.S32_LE:
+		bitdepth = 32
+	default:
+		return OpenError(fmt.Errorf("unsupported sample bits %v", d.BitDepth))
+	}
+	d.l.Log(logger.Debug, pkg+"alsa device bit depth set", "bitdepth", bitdepth)
+
+	// A 50ms period is a sensible value for low-ish latency. (this could be made configurable if needed)
+	// Some devices only accept even period sizes while others want powers of 2.
+	// So we will find the closest power of 2 to the desired period size.
+	const wantPeriod = 0.05 //seconds
+	bytesPerSecond := rate * channels * (bitdepth / 8)
+	wantPeriodSize := int(float64(bytesPerSecond) * wantPeriod)
+	nearWantPeriodSize := nearestPowerOfTwo(wantPeriodSize)
+
+	// At least two period sizes should fit within the buffer.
+	bufSize, err := d.dev.NegotiateBufferSize(nearWantPeriodSize * 2)
+	if err != nil {
+		return OpenError(err)
+	}
+	d.l.Log(logger.Debug, pkg+"alsa device buffer size set", "buffersize", bufSize)
+
+	if err = d.dev.Prepare(); err != nil {
+		return OpenError(err)
+	}
+
+	d.l.Log(logger.Debug, pkg+"successfully negotiated ALSA params")
+	return nil
+}
+
+// input continously records audio and writes it to the ringbuffer.
+// Re-opens the device and tries again if ASLA returns an error.
+func (d *Device) input() {
+	for {
+		// Check mode.
+		d.mu.Lock()
+		mode := d.mode
+		d.mu.Unlock()
+		switch mode {
+		case paused:
+			time.Sleep(time.Duration(d.RecPeriod) * time.Second)
+			continue
+		case stopped:
+			if d.dev != nil {
+				d.l.Log(logger.Debug, pkg+"closing audio device", "title", d.title)
+				d.dev.Close()
+				d.dev = nil
+			}
+			return
+		}
+
+		// Read from audio device.
+		d.l.Log(logger.Debug, pkg+"recording audio for period", "seconds", d.RecPeriod)
+		err := d.dev.Read(d.ab.Data)
+		if err != nil {
+			d.l.Log(logger.Debug, pkg+"read failed", "error", err.Error())
+			err = d.open() // re-open
+			if err != nil {
+				d.l.Log(logger.Fatal, pkg+"reopening device failed", "error", err.Error())
+				return
+			}
+			continue
+		}
+
+		// Process audio.
+		d.l.Log(logger.Debug, pkg+"processing audio")
+		toWrite := d.formatBuffer()
+
+		// Write audio to ringbuffer.
+		n, err := d.rb.Write(toWrite.Data)
+		switch err {
+		case nil:
+			d.l.Log(logger.Debug, pkg+"wrote audio to ringbuffer", "length", n)
+		case ring.ErrDropped:
+			d.l.Log(logger.Warning, pkg+"old audio data overwritten")
+		default:
+			d.l.Log(logger.Error, pkg+"unexpected ringbuffer error", "error", err.Error())
+			return
+		}
+	}
+}
+
+// Read reads from the ringbuffer, returning the number of bytes read upon success.
+func (d *Device) Read(p []byte) (int, error) {
+	// Ready ringbuffer for read.
+	_, err := d.rb.Next(rbNextTimeout)
+	if err != nil {
+		return 0, err
+	}
+
+	// Read from ring buffer.
+	return d.rb.Read(p)
+}
+
+// formatBuffer returns audio that has been converted to the desired format.
+func (d *Device) formatBuffer() alsa.Buffer {
+	var err error
+
+	// If nothing needs to be changed, return the original.
+	if d.ab.Format.Channels == d.Channels && d.ab.Format.Rate == d.SampleRate {
+		return d.ab
+	}
+	var formatted alsa.Buffer
+	if d.ab.Format.Channels != d.Channels {
+		// Convert channels.
+		// TODO(Trek): Make this work for conversions other than stereo to mono.
+		if d.ab.Format.Channels == 2 && d.Channels == 1 {
+			formatted, err = pcm.StereoToMono(d.ab)
+			if err != nil {
+				d.l.Log(logger.Fatal, pkg+"channel conversion failed", "error", err.Error())
+			}
+		}
+	}
+
+	if d.ab.Format.Rate != d.SampleRate {
+		// Convert rate.
+		formatted, err = pcm.Resample(formatted, d.SampleRate)
+		if err != nil {
+			d.l.Log(logger.Fatal, pkg+"rate conversion failed", "error", err.Error())
+		}
+	}
+
+	switch d.Codec {
+	case codecutil.PCM:
+	case codecutil.ADPCM:
+		b := bytes.NewBuffer(make([]byte, 0, adpcm.EncBytes(len(formatted.Data))))
+		enc := adpcm.NewEncoder(b)
+		_, err = enc.Write(formatted.Data)
+		if err != nil {
+			d.l.Log(logger.Fatal, pkg+"unable to encode", "error", err.Error())
+		}
+		formatted.Data = b.Bytes()
+	default:
+		d.l.Log(logger.Error, pkg+"unhandled audio codec")
+	}
+
+	return formatted
+}
+
+// nearestPowerOfTwo finds and returns the nearest power of two to the given integer.
+// If the lower and higher power of two are the same distance, it returns the higher power.
+// For negative values, 1 is returned.
+// Source: https://stackoverflow.com/a/45859570
+func nearestPowerOfTwo(n int) int {
+	if n <= 0 {
+		return 1
+	}
+	if n == 1 {
+		return 2
+	}
+	v := n
+	v--
+	v |= v >> 1
+	v |= v >> 2
+	v |= v >> 4
+	v |= v >> 8
+	v |= v >> 16
+	v++         // higher power of 2
+	x := v >> 1 // lower power of 2
+	if (v - n) > (n - x) {
+		return x
+	}
+	return v
+}

input/audio/audio_test.go

@@ -0,0 +1,105 @@
+/*
+NAME
+  audio_test.go
+
+AUTHOR
+  Trek Hopton <trek@ausocean.org>
+
+LICENSE
+  This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
+
+  It is free software: you can redistribute it and/or modify them
+  under the terms of the GNU General Public License as published by the
+  Free Software Foundation, either version 3 of the License, or (at your
+  option) any later version.
+
+  It is distributed in the hope that it will be useful, but WITHOUT
+  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+  for more details.
+
+  You should have received a copy of the GNU General Public License in gpl.txt.
+  If not, see [GNU licenses](http://www.gnu.org/licenses).
+*/
+
+package audio
+
+import (
+	"io/ioutil"
+	"os"
+	"strconv"
+	"testing"
+	"time"
+
+	"bitbucket.org/ausocean/av/codec/codecutil"
+	"bitbucket.org/ausocean/utils/logger"
+)
+
+func TestDevice(t *testing.T) {
+	// We want to open a device with a standard configuration.
+	ac := &Config{
+		SampleRate: 8000,
+		Channels:   1,
+		RecPeriod:  0.3,
+		BitDepth:   16,
+		Codec:      codecutil.ADPCM,
+	}
+	n := 2 // Number of periods to wait while recording.
+
+	// Create a new audio Device, start, read/lex, and then stop it.
+	l := logger.New(logger.Debug, os.Stderr)
+	ai, err := NewDevice(ac, l)
+	// If there was an error opening the device, skip this test.
+	if _, ok := err.(OpenError); ok {
+		t.Skip(err)
+	}
+	// For any other error, report it.
+	if err != nil {
+		t.Error(err)
+	}
+	err = ai.Start()
+	if err != nil {
+		t.Error(err)
+	}
+	chunkSize := ai.ChunkSize()
+	lexer := codecutil.NewByteLexer(&chunkSize)
+	go lexer.Lex(ioutil.Discard, ai, time.Duration(ac.RecPeriod*float64(time.Second)))
+	time.Sleep(time.Duration(ac.RecPeriod*float64(time.Second)) * time.Duration(n))
+	ai.Stop()
+}
+
+var powerTests = []struct {
+	in  int
+	out int
+}{
+	{36, 32},
+	{47, 32},
+	{3, 4},
+	{46, 32},
+	{7, 8},
+	{2, 2},
+	{36, 32},
+	{757, 512},
+	{2464, 2048},
+	{18980, 16384},
+	{70000, 65536},
+	{8192, 8192},
+	{2048, 2048},
+	{65536, 65536},
+	{-2048, 1},
+	{-127, 1},
+	{-1, 1},
+	{0, 1},
+	{1, 2},
+}
+
+func TestNearestPowerOfTwo(t *testing.T) {
+	for _, tt := range powerTests {
+		t.Run(strconv.Itoa(tt.in), func(t *testing.T) {
+			v := nearestPowerOfTwo(tt.in)
+			if v != tt.out {
+				t.Errorf("got %v, want %v", v, tt.out)
+			}
+		})
+	}
+}

revid/config.go

@@ -2,11 +2,9 @@
 NAME
   Config.go
 
-DESCRIPTION
-  See Readme.md
-
 AUTHORS
   Saxon A. Nelson-Milton <saxon@ausocean.org>
+  Trek Hopton <trek@ausocean.org>
 
 LICENSE
   Config.go is Copyright (C) 2017-2018 the Australian Ocean Lab (AusOcean)
@@ -30,6 +28,7 @@ package revid
 import (
 	"errors"
 
+	"bitbucket.org/ausocean/av/codec/codecutil"
 	"bitbucket.org/ausocean/utils/logger"
 )
 
@@ -75,6 +74,7 @@ const (
 	Raspivid
 	V4L
 	RTSP
+	Audio
 
 	// Outputs.
 	RTMP
@@ -93,6 +93,7 @@ const (
 	defaultInput              = Raspivid
 	defaultOutput             = HTTP
 	defaultFrameRate          = 25
+	defaultWriteRate          = 25
 	defaultWidth              = 1280
 	defaultHeight             = 720
 	defaultIntraRefreshPeriod = 100
@@ -101,7 +102,7 @@ const (
 	defaultBitrate            = 400000
 	defaultFramesPerClip      = 1
 	httpFramesPerClip         = 560
-	defaultInputCodec         = H264
+	defaultInputCodec         = codecutil.H264
 	defaultVerbosity          = logger.Error
 	defaultRtpAddr            = "localhost:6970"
 	defaultBurstPeriod        = 10 // Seconds
@@ -109,6 +110,12 @@ const (
 	defaultBrightness         = 50
 	defaultExposure           = "auto"
 	defaultAutoWhiteBalance   = "auto"
+
+	defaultAudioInputCodec = codecutil.ADPCM
+	defaultSampleRate      = 48000
+	defaultBitDepth        = 16
+	defaultChannels        = 1
+	defaultRecPeriod       = 1.0
 )
 
 // Config provides parameters relevant to a revid instance. A new config must
@@ -183,6 +190,9 @@ type Config struct {
 	// Raspivid input supports custom framerate.
 	FrameRate uint
 
+	// WriteRate is how many times a second revid encoders will be written to.
+	WriteRate float64
+
 	// HTTPAddress defines a custom HTTP destination if we do not wish to use that
 	// defined in /etc/netsender.conf.
 	HTTPAddress string
@@ -215,6 +225,13 @@ type Config struct {
 	// defined at the start of the file.
 	AutoWhiteBalance string
 
+	// Audio
+	SampleRate int     // Samples a second (Hz).
+	RecPeriod  float64 // How many seconds to record at a time.
+	Channels   int     // Number of audio channels, 1 for mono, 2 for stereo.
+	BitDepth   int     // Sample bit depth.
+	ChunkSize  int     // ChunkSize is the size of the chunks in the audio.Device's ringbuffer.
+
 	RTPAddress     string // RTPAddress defines the RTP output destination.
 	BurstPeriod    uint   // BurstPeriod defines the revid burst period in seconds.
 	Rotation       uint   // Rotation defines the video rotation angle in degrees Raspivid input.
@@ -240,7 +257,7 @@ func (c *Config) Validate(r *Revid) error {
 	}
 
 	switch c.Input {
-	case Raspivid, V4L, File, RTSP:
+	case Raspivid, V4L, File, Audio, RTSP:
 	case NothingDefined:
 		c.Logger.Log(logger.Info, pkg+"no input type defined, defaulting", "input", defaultInput)
 		c.Input = defaultInput
@@ -249,7 +266,7 @@ func (c *Config) Validate(r *Revid) error {
 	}
 
 	switch c.InputCodec {
-	case H264:
+	case codecutil.H264:
 		// FIXME(kortschak): This is not really what we want.
 		// Configuration really needs to be rethought here.
 		if c.Quantize && c.Quantization == 0 {
@@ -260,18 +277,22 @@ func (c *Config) Validate(r *Revid) error {
 			return errors.New("bad bitrate and quantization combination for H264 input")
 		}
 
-	case MJPEG:
+	case codecutil.MJPEG:
 		if c.Quantization > 0 || c.Bitrate == 0 {
 			return errors.New("bad bitrate or quantization for mjpeg input")
 		}
-
+	case codecutil.PCM, codecutil.ADPCM:
-	case NothingDefined:
-		c.Logger.Log(logger.Info, pkg+"no input codec defined, defaulting", "inputCodec", defaultInputCodec)
-		c.InputCodec = defaultInputCodec
-		c.Logger.Log(logger.Info, pkg+"defaulting quantization", "quantization", defaultQuantization)
-		c.Quantization = defaultQuantization
 	default:
-		return errors.New("bad input codec defined in config")
+		switch c.Input {
+		case Audio:
+			c.Logger.Log(logger.Info, pkg+"input is audio but no codec defined, defaulting", "inputCodec", defaultAudioInputCodec)
+			c.InputCodec = defaultAudioInputCodec
+		default:
+			c.Logger.Log(logger.Info, pkg+"no input codec defined, defaulting", "inputCodec", defaultInputCodec)
+			c.InputCodec = defaultInputCodec
+			c.Logger.Log(logger.Info, pkg+"defaulting quantization", "quantization", defaultQuantization)
+			c.Quantization = defaultQuantization
+		}
 	}
 
 	if c.Outputs == nil {
@@ -330,6 +351,31 @@ func (c *Config) Validate(r *Revid) error {
 		c.FrameRate = defaultFrameRate
 	}
 
+	if c.SampleRate == 0 {
+		c.Logger.Log(logger.Info, pkg+"no sample rate defined, defaulting", "sampleRate", defaultSampleRate)
+		c.SampleRate = defaultSampleRate
+	}
+
+	if c.Channels == 0 {
+		c.Logger.Log(logger.Info, pkg+"no number of channels defined, defaulting", "Channels", defaultChannels)
+		c.Channels = defaultChannels
+	}
+
+	if c.BitDepth == 0 {
+		c.Logger.Log(logger.Info, pkg+"no bit depth defined, defaulting", "BitDepth", defaultBitDepth)
+		c.BitDepth = defaultBitDepth
+	}
+
+	if c.RecPeriod == 0 {
+		c.Logger.Log(logger.Info, pkg+"no record period defined, defaulting", "recPeriod", defaultRecPeriod)
+		c.RecPeriod = defaultRecPeriod
+	}
+
+	if c.WriteRate == 0 {
+		c.Logger.Log(logger.Info, pkg+"no write rate defined, defaulting", "writeRate", defaultWriteRate)
+		c.WriteRate = defaultWriteRate
+	}
+
 	if c.Bitrate == 0 {
 		c.Logger.Log(logger.Info, pkg+"no bitrate defined, defaulting", "bitrate", defaultBitrate)
 		c.Bitrate = defaultBitrate

revid/revid.go

@@ -2,13 +2,11 @@
 NAME
   revid.go
 
-DESCRIPTION
-  See Readme.md
-
 AUTHORS
   Saxon A. Nelson-Milton <saxon@ausocean.org>
   Alan Noble <alan@ausocean.org>
   Dan Kortschak <dan@ausocean.org>
+  Trek Hopton <trek@ausocean.org>
 
 LICENSE
   revid is Copyright (C) 2017-2018 the Australian Ocean Lab (AusOcean)
@@ -27,6 +25,7 @@ LICENSE
   in gpl.txt.  If not, see http://www.gnu.org/licenses.
 */
 
+// Package revid provides an API for reading, transcoding, and writing audio/video streams and files.
 package revid
 
 import (
@@ -41,10 +40,12 @@ import (
 	"sync"
 	"time"
 
+	"bitbucket.org/ausocean/av/codec/codecutil"
 	"bitbucket.org/ausocean/av/codec/h264"
 	"bitbucket.org/ausocean/av/codec/h265"
 	"bitbucket.org/ausocean/av/container/flv"
 	"bitbucket.org/ausocean/av/container/mts"
+	"bitbucket.org/ausocean/av/input/audio"
 	"bitbucket.org/ausocean/av/protocol/rtcp"
 	"bitbucket.org/ausocean/av/protocol/rtp"
 	"bitbucket.org/ausocean/av/protocol/rtsp"
@@ -173,13 +174,15 @@ func (r *Revid) reset(config Config) error {
 	r.config.Logger.SetLevel(config.LogLevel)
 
 	err = r.setupPipeline(
-		func(dst io.WriteCloser, fps int) (io.WriteCloser, error) {
+		func(dst io.WriteCloser, fps float64) (io.WriteCloser, error) {
 			var st int
 			switch r.config.Input {
 			case Raspivid, File, V4L:
 				st = mts.EncodeH264
 			case RTSP:
 				st = mts.EncodeH265
+			case Audio:
+				st = mts.EncodeAudio
 			}
 			e := mts.NewEncoder(dst, float64(fps), st)
 			return e, nil
@@ -215,7 +218,7 @@ func (r *Revid) setConfig(config Config) error {
 // mtsEnc and flvEnc will be called to obtain an mts encoder and flv encoder
 // respectively. multiWriter will be used to create an ioext.multiWriteCloser
 // so that encoders can write to multiple senders.
-func (r *Revid) setupPipeline(mtsEnc, flvEnc func(dst io.WriteCloser, rate int) (io.WriteCloser, error), multiWriter func(...io.WriteCloser) io.WriteCloser) error {
+func (r *Revid) setupPipeline(mtsEnc func(dst io.WriteCloser, rate float64) (io.WriteCloser, error), flvEnc func(dst io.WriteCloser, rate int) (io.WriteCloser, error), multiWriter func(...io.WriteCloser) io.WriteCloser) error {
 	// encoders will hold the encoders that are required for revid's current
 	// configuration.
 	var encoders []io.WriteCloser
@@ -259,7 +262,7 @@ func (r *Revid) setupPipeline(mtsEnc, flvEnc func(dst io.WriteCloser, rate int)
 	// as a destination.
 	if len(mtsSenders) != 0 {
 		mw := multiWriter(mtsSenders...)
-		e, _ := mtsEnc(mw, int(r.config.FrameRate))
+		e, _ := mtsEnc(mw, r.config.WriteRate)
 		encoders = append(encoders, e)
 	}
 
@@ -289,6 +292,9 @@ func (r *Revid) setupPipeline(mtsEnc, flvEnc func(dst io.WriteCloser, rate int)
 	case RTSP:
 		r.setupInput = r.startRTSPCamera
 		r.lexTo = h265.NewLexer(false).Lex
+	case Audio:
+		r.setupInput = r.startAudioDevice
+		r.lexTo = codecutil.NewByteLexer(&r.config.ChunkSize).Lex
 	}
 
 	return nil
@@ -533,7 +539,7 @@ func (r *Revid) startRaspivid() (func() error, error) {
 	switch r.config.InputCodec {
 	default:
 		return nil, fmt.Errorf("revid: invalid input codec: %v", r.config.InputCodec)
-	case H264:
+	case codecutil.H264:
 		args = append(args,
 			"--codec", "H264",
 			"--inline",
@@ -542,7 +548,7 @@ func (r *Revid) startRaspivid() (func() error, error) {
 		if r.config.Quantize {
 			args = append(args, "-qp", fmt.Sprint(r.config.Quantization))
 		}
-	case MJPEG:
+	case codecutil.MJPEG:
 		args = append(args, "--codec", "MJPEG")
 	}
 	r.config.Logger.Log(logger.Info, pkg+"raspivid args", "raspividArgs", strings.Join(args, " "))
@@ -615,10 +621,55 @@ func (r *Revid) setupInputForFile() (func() error, error) {
 
 	// TODO(kortschak): Maybe we want a context.Context-aware parser that we can stop.
 	r.wg.Add(1)
-	go r.processFrom(f, time.Second/time.Duration(r.config.FrameRate))
+	go r.processFrom(f, 0)
 	return func() error { return f.Close() }, nil
 }
 
+// startAudioDevice is used to start capturing audio from an audio device and processing it.
+// It returns a function that can be used to stop the device and any errors that occur.
+func (r *Revid) startAudioDevice() (func() error, error) {
+	// Create audio device.
+	ac := &audio.Config{
+		SampleRate: r.config.SampleRate,
+		Channels:   r.config.Channels,
+		RecPeriod:  r.config.RecPeriod,
+		BitDepth:   r.config.BitDepth,
+		Codec:      r.config.InputCodec,
+	}
+	mts.Meta.Add("sampleRate", strconv.Itoa(r.config.SampleRate))
+	mts.Meta.Add("channels", strconv.Itoa(r.config.Channels))
+	mts.Meta.Add("period", fmt.Sprintf("%.6f", r.config.RecPeriod))
+	mts.Meta.Add("bitDepth", strconv.Itoa(r.config.BitDepth))
+	switch r.config.InputCodec {
+	case codecutil.PCM:
+		mts.Meta.Add("codec", "pcm")
+	case codecutil.ADPCM:
+		mts.Meta.Add("codec", "adpcm")
+	default:
+		r.config.Logger.Log(logger.Fatal, pkg+"no audio codec set in config")
+	}
+
+	ai, err := audio.NewDevice(ac, r.config.Logger)
+	if err != nil {
+		r.config.Logger.Log(logger.Fatal, pkg+"failed to create audio device", "error", err.Error())
+	}
+
+	// Start audio device
+	err = ai.Start()
+	if err != nil {
+		r.config.Logger.Log(logger.Fatal, pkg+"failed to start audio device", "error", err.Error())
+	}
+
+	// Process output from audio device.
+	r.config.ChunkSize = ai.ChunkSize()
+	r.wg.Add(1)
+	go r.processFrom(ai, time.Duration(float64(time.Second)/r.config.WriteRate))
+	return func() error {
+		ai.Stop()
+		return nil
+	}, nil
+}
+
 // startRTSPCamera uses RTSP to request an RTP stream from an IP camera. An RTP
 // client is created from which RTP packets containing either h264/h265 can read
 // by the selected lexer.

revid/revid_test.go

@@ -41,7 +41,7 @@ import (
 const raspividPath = "/usr/local/bin/raspivid"
 
 // Suppress all test logging, except for t.Errorf output.
-var silent bool
+var silent = true
 
 // TestRaspivid tests that raspivid starts correctly.
 // It is intended to be run on a Raspberry Pi.
@@ -232,7 +232,7 @@ func TestResetEncoderSenderSetup(t *testing.T) {
 
 		// This logic is what we want to check.
 		err = rv.setupPipeline(
-			func(dst io.WriteCloser, rate int) (io.WriteCloser, error) {
+			func(dst io.WriteCloser, rate float64) (io.WriteCloser, error) {
 				return &tstMtsEncoder{dst: dst}, nil
 			},
 			func(dst io.WriteCloser, rate int) (io.WriteCloser, error) {

revid/senders.go

@@ -57,7 +57,7 @@ type httpSender struct {
 	log    func(lvl int8, msg string, args ...interface{})
 }
 
-// newMinimalHttpSender returns a pointer to a new minimalHttpSender.
+// newHttpSender returns a pointer to a new httpSender.
 func newHttpSender(ns *netsender.Sender, log func(lvl int8, msg string, args ...interface{})) *httpSender {
 	return &httpSender{
 		client: ns,