81 lines
2.5 KiB
Go
81 lines
2.5 KiB
Go
|
package audio
|
||
|
|
|
||
|
|
import (
|
||
|
|
"encoding/binary"
|
||
|
|
"fmt"
|
||
|
|
)
|
||
|
|
|
||
|
|
// pcmFormat beschreibt rohe PCM-Samples, wie sie oto zur Wiedergabe
|
||
|
|
// braucht.
|
||
|
|
type pcmFormat struct {
|
||
|
|
SampleRate int
|
||
|
|
Channels int
|
||
|
|
BitsPerCoded int // 8 (unsigned), 16 (signed LE), 32 (nur IEEE float)
|
||
|
|
Float bool
|
||
|
|
}
|
||
|
|
|
||
|
|
// decodeWAV zerlegt RIFF/WAVE-Bytes in Format und PCM-Daten.
|
||
|
|
//
|
||
|
|
// Unterstützt: PCM (AudioFormat 1) mit 8/16 Bit sowie IEEE-Float
|
||
|
|
// (AudioFormat 3) mit 32 Bit; Mono und Stereo. Alles andere — erst recht
|
||
|
|
// komprimierte Formate — ist ein klarer Fehler; die OS-TTS-Wege von
|
||
|
|
// osbuiltin liefern ausschließlich 16-Bit-PCM-WAV.
|
||
|
|
func decodeWAV(data []byte) (pcmFormat, []byte, error) {
|
||
|
|
var f pcmFormat
|
||
|
|
if len(data) < 12 {
|
||
|
|
return f, nil, fmt.Errorf("wav: zu kurz für RIFF-Header (%d Bytes)", len(data))
|
||
|
|
}
|
||
|
|
if string(data[0:4]) != "RIFF" || string(data[8:12]) != "WAVE" {
|
||
|
|
return f, nil, fmt.Errorf("wav: kein RIFF/WAVE-Header")
|
||
|
|
}
|
||
|
|
|
||
|
|
var haveFmt bool
|
||
|
|
var payload []byte
|
||
|
|
pos := 12
|
||
|
|
for pos+8 <= len(data) {
|
||
|
|
id := string(data[pos : pos+4])
|
||
|
|
size := int(binary.LittleEndian.Uint32(data[pos+4 : pos+8]))
|
||
|
|
body := pos + 8
|
||
|
|
if body+size > len(data) {
|
||
|
|
return f, nil, fmt.Errorf("wav: Chunk %q überragt die Datei (Size %d)", id, size)
|
||
|
|
}
|
||
|
|
switch id {
|
||
|
|
case "fmt ":
|
||
|
|
if size < 16 {
|
||
|
|
return f, nil, fmt.Errorf("wav: fmt-Chunk zu kurz (%d)", size)
|
||
|
|
}
|
||
|
|
audioFormat := binary.LittleEndian.Uint16(data[body : body+2])
|
||
|
|
f.Channels = int(binary.LittleEndian.Uint16(data[body+2 : body+4]))
|
||
|
|
f.SampleRate = int(binary.LittleEndian.Uint32(data[body+4 : body+8]))
|
||
|
|
f.BitsPerCoded = int(binary.LittleEndian.Uint16(data[body+14 : body+16]))
|
||
|
|
switch {
|
||
|
|
case audioFormat == 1 && (f.BitsPerCoded == 8 || f.BitsPerCoded == 16):
|
||
|
|
case audioFormat == 3 && f.BitsPerCoded == 32:
|
||
|
|
f.Float = true
|
||
|
|
default:
|
||
|
|
return f, nil, fmt.Errorf("wav: nicht unterstütztes Format (AudioFormat %d, %d Bit) — erlaubt: PCM 8/16 Bit, Float 32 Bit",
|
||
|
|
audioFormat, f.BitsPerCoded)
|
||
|
|
}
|
||
|
|
if f.Channels < 1 || f.Channels > 2 {
|
||
|
|
return f, nil, fmt.Errorf("wav: %d Kanäle (erlaubt: 1, 2)", f.Channels)
|
||
|
|
}
|
||
|
|
if f.SampleRate <= 0 {
|
||
|
|
return f, nil, fmt.Errorf("wav: SampleRate %d", f.SampleRate)
|
||
|
|
}
|
||
|
|
haveFmt = true
|
||
|
|
case "data":
|
||
|
|
payload = data[body : body+size]
|
||
|
|
}
|
||
|
|
pos = body + size
|
||
|
|
if size%2 == 1 { // Chunks sind word-aligned
|
||
|
|
pos++
|
||
|
|
}
|
||
|
|
}
|
||
|
|
if !haveFmt {
|
||
|
|
return f, nil, fmt.Errorf("wav: kein fmt-Chunk")
|
||
|
|
}
|
||
|
|
if payload == nil {
|
||
|
|
return f, nil, fmt.Errorf("wav: kein data-Chunk")
|
||
|
|
}
|
||
|
|
return f, payload, nil
|
||
|
|
}
|