drwily/Pete
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add server-side Piper read-aloud with voice picker

Browse Source 
Reader read-aloud now streams neural WAV audio from a new POST /api/tts
endpoint that shells out to Piper, instead of the browser's Web Speech
voice. Each paragraph is synthesized on demand with the next one
prefetched during playback, keeping the existing highlight/scroll sync.

Voices are configured under [web.tts] (piper binary + voices_dir + a
labelled voice list) and exposed to the client as window.PETE_TTS; the
reader gets a Voice selector in the Aa menu, persisted per-device. Still
a signed-in-only perk and gated on auth.
This commit is contained in:
prosolis
2026-07-07 23:00:27 -07:00
parent fceeb12ad5
commit dbcb459908
9 changed files with 436 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
internal/config/config.go
View File

@@ -28,6 +28,7 @@ type WebConfig struct {
BaseURL string `toml:"base_url"` // public URL (used in metadata only)
Auth AuthConfig `toml:"auth"` // optional OIDC sign-in (Authentik)
Push PushConfig `toml:"push"` // optional Web Push digests (VAPID)
TTS TTSConfig `toml:"tts"` // optional server-side neural read-aloud (Piper)
// AdminSubs is the allowlist of OIDC subjects allowed to view the
// owner-facing source-health dashboard at /status. Empty means the page is
// inaccessible to everyone (returns 404). Requires auth to be enabled.
@@ -53,6 +54,29 @@ type PushConfig struct {
MinStories int `toml:"min_stories"`
}
// TTSConfig wires server-side neural read-aloud (Piper). When enabled,
// signed-in users get the reader's "Listen" button backed by real Piper voices
// instead of the browser's robotic Web Speech voice. Read-aloud is a signed-in
// perk, so this does nothing unless auth is also enabled.
type TTSConfig struct {
Enabled bool `toml:"enabled"`
PiperBin string `toml:"piper_bin"` // path to the piper executable
VoicesDir string `toml:"voices_dir"` // directory holding <id>.onnx (+ .onnx.json) models
// Voices lists the voices to offer, in menu order. Each id is a model
// filename stem, so id "en_US-ryan-high" maps to <voices_dir>/en_US-ryan-high.onnx.
// Leave empty to auto-discover every *.onnx in voices_dir.
Voices []VoiceConfig `toml:"voices"`
// Default is the voice id selected until the reader picks another. Empty
// falls back to the first available voice.
Default string `toml:"default"`
}
// VoiceConfig is one selectable Piper voice.
type VoiceConfig struct {
ID string `toml:"id"` // model filename stem, e.g. "en_US-ryan-high"
Label string `toml:"label"` // menu label, e.g. "Ryan — US, male"
}
// AuthConfig wires Pete's web UI to an OIDC provider (our Authentik instance).
// When enabled, signed-in users get their preferences stored server-side keyed
// by the OIDC subject; anonymous visitors keep using browser localStorage.