Files
Maven/cmd/mavttsd/piper_handler.go
T

130 lines
3.0 KiB
Go

package main
import (
"bytes"
"context"
"encoding/binary"
"fmt"
"io"
"os/exec"
"github.com/kami/maven/internal/audio"
"github.com/kami/maven/internal/worker"
)
type piperHandler struct {
piperPath string
modelPath string
configPath string
espeakData string
tashkeelModel string
}
func newPiperHandler(piperPath, modelPath, espeakData, tashkeelModel string) *piperHandler {
return &piperHandler{
piperPath: piperPath,
modelPath: modelPath,
configPath: modelPath + ".json",
espeakData: espeakData,
tashkeelModel: tashkeelModel,
}
}
func (h *piperHandler) Synthesize(ctx context.Context, req worker.SynthesizeReq) (worker.SynthesizeResp, error) {
var stderr bytes.Buffer
args := []string{
"--model", h.modelPath,
"--config", h.configPath,
"--output_raw",
"--quiet",
}
if h.espeakData != "" {
args = append(args, "--espeak_data", h.espeakData)
}
if h.tashkeelModel != "" {
args = append(args, "--tashkeel_model", h.tashkeelModel)
}
cmd := exec.CommandContext(ctx, h.piperPath, args...)
cmd.Stderr = &stderr
stdin, err := cmd.StdinPipe()
if err != nil {
return worker.SynthesizeResp{}, fmt.Errorf("piper: stdin pipe: %w", err)
}
stdout, err := cmd.StdoutPipe()
if err != nil {
return worker.SynthesizeResp{}, fmt.Errorf("piper: stdout pipe: %w", err)
}
if err := cmd.Start(); err != nil {
return worker.SynthesizeResp{}, fmt.Errorf("piper: start: %w", err)
}
_, _ = io.WriteString(stdin, req.Text)
stdin.Close()
rawPCM, readErr := io.ReadAll(stdout)
waitErr := cmd.Wait()
if waitErr != nil {
errMsg := stderr.String()
if errMsg != "" {
return worker.SynthesizeResp{}, fmt.Errorf("piper: %s: %s", waitErr, errMsg)
}
return worker.SynthesizeResp{}, fmt.Errorf("piper: %w", waitErr)
}
if readErr != nil {
return worker.SynthesizeResp{}, fmt.Errorf("piper: read stdout: %w", readErr)
}
if len(rawPCM) == 0 {
return worker.SynthesizeResp{}, fmt.Errorf("piper: no audio output")
}
resampled := resample22050To16000(rawPCM)
return worker.SynthesizeResp{
Audio: audio.Audio{
Format: audio.PCM16kMono,
Bytes: resampled,
},
}, nil
}
// resample22050To16000 converts raw 16-bit PCM from 22050 Hz to 16000 Hz
// using linear interpolation.
func resample22050To16000(input []byte) []byte {
if len(input) < 2 {
return nil
}
nSamples := len(input) / 2
outSamples := int(float64(nSamples) * 16000.0 / 22050.0)
output := make([]byte, outSamples*2)
ratio := 22050.0 / 16000.0
for i := 0; i < outSamples; i++ {
srcPos := float64(i) * ratio
srcIdx := int(srcPos)
frac := srcPos - float64(srcIdx)
if srcIdx >= nSamples-1 {
v := int16(binary.LittleEndian.Uint16(input[(nSamples-1)*2:]))
binary.LittleEndian.PutUint16(output[i*2:], uint16(v))
continue
}
v0 := int16(binary.LittleEndian.Uint16(input[srcIdx*2:]))
v1 := int16(binary.LittleEndian.Uint16(input[(srcIdx+1)*2:]))
interpolated := int16(float64(v0)*(1-frac) + float64(v1)*frac)
binary.LittleEndian.PutUint16(output[i*2:], uint16(interpolated))
}
return output
}