github.com/steipete/sag

Fixed

• Explicit --format values now override output-extension inference, preserving the requested ElevenLabs audio format. (#21, #22, thanks @100yenadmin)

Verification

• Release commit: 4797496
• Exact-head CI: https://github.com/steipete/sag/actions/runs/28507322673
• Docs deployment: https://github.com/steipete/sag/actions/runs/28507322694
• Authenticated live proof: an explicit 192 kbps MP3 format overrode a conflicting .wav output extension; the generated audio was independently detected as MP3 with positive duration. No playback, private text, or persistent settings were used.
• Source-blind behavior validation: 5 passed, 0 failed, 0 blocked.
• Checksums: https://github.com/steipete/sag/releases/download/v0.4.1/sag_0.4.1_checksums.txt

Fixed

• Explicit --format values now override output-extension inference, preserving the requested ElevenLabs audio format. (#21/#22, thanks @100yenadmin)

Added

• Added live-tested 60db provider support via /default-voices, /myvoices, and /tts-synthesize, including bounded NDJSON decoding, incomplete-response rejection, and PCM-to-WAV playback. (steipete/sag#20, thanks @manishEMS47)

Changed

• Release archives now include target-specific macOS and Linux assets for Homebrew and aqua installers.

Verification

• Release candidate CI: https://github.com/steipete/sag/actions/runs/27319088377
• Docs deployment: https://github.com/steipete/sag/actions/runs/27319088365
• Final release workflow: https://github.com/steipete/sag/actions/runs/27319436814
• Homebrew formula update: https://github.com/steipete/homebrew-tap/actions/runs/27319495324
• Checksums: https://github.com/steipete/sag/releases/download/v0.4.0/sag_0.4.0_checksums.txt
• Authenticated 60db proof: the Homebrew-installed sag 0.4.0 listed live voices, synthesized through the real provider, produced a 48 kHz mono PCM/WAV file, and completed oto playback.

Added

• Added live-tested 60db provider support via /default-voices, /myvoices, and /tts-synthesize, including bounded NDJSON decoding, incomplete-response rejection, and PCM-to-WAV playback. (#20, thanks @manishEMS47)

Changed

• Release archives now include target-specific macOS and Linux assets for Homebrew and aqua installers.

Added

• Voice discovery: semantic --query over name/description/labels, repeatable --label filters, preview playback via --try, metadata caching, and server-side name search when supported.
• Bare sag now reads piped stdin like macOS say, so echo "hello" | sag speaks the text without requiring the speak subcommand. (#14, thanks @atdrendel)

Fixed

• --format and .ogg/.opus output paths now request native ElevenLabs Opus output instead of always receiving MP3. (#16, thanks @derspotter)
• Long TTS generation no longer hits hidden 60s/90s internal timeouts; use --timeout or SAG_TIMEOUT when you want sag to enforce a limit. (#17/#18, thanks @sumurtk2)
• --no-play and --no-stream are now accepted aliases for --play=false and --stream=false.
• macOS playback now defaults to afplay for better AirPlay routing, with --player oto / SAG_PLAYER=oto available as a fallback. (#15, thanks @s0up4200)

Added

• Voice discovery: semantic --query over name/description/labels, repeatable --label filters, preview playback via --try, metadata caching, and server-side name search when supported.
• Bare sag now reads piped stdin like macOS say, so echo "hello" | sag speaks the text without requiring the speak subcommand. (#14, thanks @atdrendel)

Fixed

• --format and .ogg/.opus output paths now request native ElevenLabs Opus output instead of always receiving MP3. (#16, thanks @derspotter)
• Long TTS generation no longer hits hidden 60s/90s internal timeouts; use --timeout or SAG_TIMEOUT when you want sag to enforce a limit. (#17/#18, thanks @sumurtk2)
• --no-play and --no-stream are now accepted aliases for --play=false and --stream=false.
• macOS playback now defaults to afplay for better AirPlay routing, with --player oto / SAG_PLAYER=oto available as a fallback. (#15, thanks @s0up4200)

Fixed

• Voice ID resolution respects --voice-id and avoids misclassifying long names; --rate now overrides --speed validation. (#7, thanks @joelbdavies)
• Voice name matching now uses exact/substring checks without falling back to unrelated voices; voice search is handled client-side. (#8, thanks @joelbdavies)
• sag help and sag completion no longer default to speak. (#4, thanks @GiGurra)

Added

• --api-key-file and ELEVENLABS_API_KEY_FILE/SAG_API_KEY_FILE support for loading API keys from a file. (#6, thanks @GiGurra)

Fixed

• Voice ID resolution respects --voice-id and avoids misclassifying long names; --rate now overrides --speed validation. (#7, thanks @joelbdavies)
• Voice name matching now uses exact/substring checks without falling back to unrelated voices; voice search is handled client-side. (#8, thanks @joelbdavies)
• sag help and sag completion no longer default to speak. (#4, thanks @GiGurra)

Added

• --api-key-file and ELEVENLABS_API_KEY_FILE/SAG_API_KEY_FILE support for loading API keys from a file. (#6, thanks @GiGurra)

Fixed

• -o/--output now disables speaker playback by default unless --play is explicitly set.

Fixed

• -o/--output now disables speaker playback by default unless --play is explicitly set. Previously -o saved to file AND played through speakers, which was confusing.

Added

• Voice control flags: --stability, --similarity/--similarity-boost, --style, --speaker-boost/--no-speaker-boost.
• Request controls: --seed, --normalize (auto|on|off), --lang (ISO 639-1).
• --metrics to print basic request stats (chars/bytes/duration) to stderr.
• sag prompting command + README section with prompting tips.

Changed

• Default model is now eleven_v3 (override with --model-id eleven_multilingual_v2 for a stable baseline).

Homebrew (source tarball)

• SHA256: a505a5eb0d72ede07768e96f17574c37959dacdde03434360b1a74c60d9c11cc

Added

• Voice control flags: --stability, --similarity/--similarity-boost, --style, --speaker-boost/--no-speaker-boost.
• Request controls: --seed, --normalize (auto|on|off), --lang (ISO 639-1).
• --metrics to print basic request stats (chars/bytes/duration) to stderr.
• sag prompting command and README section with prompting tips.

Changed

• Default model is now eleven_v3 (override with --model-id eleven_multilingual_v2 for a stable baseline).

docs: add sag release checklist

Added

• Initial release of sag ElevenLabs TTS CLI with macOS say-style flags.
• Streaming default playback to speakers with optional file output; cross-platform audio via go-mp3 + oto.
• Voice listing (sag voices) and voice resolution by name/ID, including ? to list.
• macOS say compatibility: -v/--voice, -r/--rate, -f/--input-file, -o/--output, plus accepted no-op flags.
• Auto-routing bare sag ... text/flags to speak subcommand, including npm/pnpm -- pass-through support.
• Speed control (--speed or --rate), latency tier selection, model/format overrides with extension inference.
• Default voice fallback to first available when none provided; env support ELEVENLABS_VOICE_ID or SAG_VOICE_ID.
• Config via ELEVENLABS_API_KEY/SAG_API_KEY; --api-key and --base-url overrides.
• Tests for format inference, text sourcing, voice resolution helpers, and default speak routing behavior.
• Help/README improvements: feature overview, examples for subcommand-less usage, and voice discovery guidance.
• Homebrew tap formula (brew install steipete/tap/sag) and release playbook.
• CI workflow (lint + tests) and golangci-lint config.
• Documentation: README, docs/spec.md, and usage examples.
• Version flag (--version / -V) reporting 0.1.0; help available without API key.