adeelahmad · Oct 10, 2026 · Oct 10, 2026 · Oct 10, 2026 · Oct 10, 2026
diff --git a/DEVLOG.md b/DEVLOG.md
@@ -2,6 +2,22 @@
 
 Plans and progress for work in flight. Newest first.
 
+## 2026-10-10 · More voices for spoken answers
+
+Goal (Adeel): read answers and podcasts aloud with ElevenLabs, macOS `say`, Google Gemini TTS or StyleTTS2, with
+the voice picked in Settings. ElevenLabs and Deepgram were already there (#116); the providers stay shared with
+podcasts through `speech.speak()`.
+
+Todo:
+
+- [x] Gemini TTS and macOS `say` as text-to-speech providers; `GET /voice/voices` lists each provider's voices;
+      `POST /voice/speak` takes a voice to try
+- [ ] StyleTTS2 as an opt-in local service (off by default; heavy)
+- [x] Settings → AI assistant → Voice: pick a voice from the list and hear it; Gemini in Speech providers
+
+Refine later: pick a second `say` voice for podcast hosts from what the Mac has; Gemini's multi-speaker mode for a
+whole podcast in one call; ElevenLabs voice settings (stability, style); cache the voice lists.
+
 ## 2026-10-10 · Podcasts from picked knowledge
 
 Goal (Adeel): pick documents, search results, topics, notes or graph concepts and get a two-host, Recast-style

diff --git a/docs/assistant.md b/docs/assistant.md
@@ -71,12 +71,13 @@ Voice goes through one hook, `useVoice()` in `nextjs-frontend/lib/voice.ts` (`li
   listening. Without a server engine, or with `voice.input: browser`, the browser's own recognition listens (Chrome,
   Edge and Safari; Chrome sends the audio to its own speech service).
 - **Speaking.** Answers are read by a speech model when one is set (`voice.tts_model`: any OpenAI-compatible
-  `/audio/speech`, such as Kokoro-FastAPI, LocalAI or OpenAI's `tts-1`, on `voice.tts_base_url` or the LLM provider);
-  else by the browser.
+  `/audio/speech`, such as Kokoro-FastAPI, LocalAI or OpenAI's `tts-1`, on `voice.tts_base_url` or the LLM provider),
+  or by ElevenLabs, Deepgram, Google Gemini or macOS `say` ([Speech providers](speech-providers.md#voices)); else by
+  the browser.
 
 Settings → AI assistant → Voice holds these. The API: `GET /voice` (what the server does; it also starts loading the
-engine), `POST /voice/transcribe` (a clip as the raw body; nothing is kept), `POST /voice/speak` (`{text}`; 204 when
-the browser should read it).
+engine), `POST /voice/transcribe` (a clip as the raw body; nothing is kept), `POST /voice/speak` (`{text, voice?}`;
+204 when the browser should read it), `GET /voice/voices` (the voices a provider reads in).
 
 ## The assistant's tools
 

diff --git a/docs/speech-providers.md b/docs/speech-providers.md
@@ -15,6 +15,8 @@ summaries, chapters and entities as optional imports, per-minute prices in the a
 | ElevenLabs | Scribe | yes | yes | no | laughter, applause, … | yes |
 | AssemblyAI | Universal, Slam-1 | yes | detected | from sentiment | no | no |
 | Deepgram | Nova | yes | detected | from sentiment (English) | no | Aura |
+| Google Gemini | no | no | no | no | no | `gemini-2.5-flash-preview-tts`, 30 prebuilt voices |
+| macOS `say` | no | no | no | no | no | the Mac's own voices; no key, nothing leaves the Mac |
 
 Sentiment becomes emotion as positive → Happy, negative → Sad, neutral → Neutral (turn it off with
 `speech.sentiment`). SenseVoice's own emotion tags stay richer; a provider only reports sentiment.
@@ -35,18 +37,33 @@ Then pick where each task goes:
 - **Speaker separation → Method**: *Speech provider* uses the speakers the provider found; *Auto* uses them when a
   transcript has some, and otherwise works as before. Voice IDs across recordings still come from local voiceprints.
 - **AI assistant → Voice**: *Engine that hears it* (default: the transcription engine) and *Spoken answers by* (an
-  OpenAI-compatible speech server, ElevenLabs or Deepgram Aura).
+  OpenAI-compatible speech server, ElevenLabs, Deepgram Aura, Google Gemini or macOS `say`) and *Voice*.
+
+## Voices
+
+`GET /voice/voices?provider=…` lists the voices a provider reads in, as `{id, name, detail}`: ElevenLabs lists the
+voices in your account (premade, cloned and from the library), Deepgram its Aura models, `say` the voices installed on
+the Mac (System Settings → Accessibility → Spoken Content adds more), and an OpenAI-compatible server its own when it
+has `/audio/voices` (Kokoro-FastAPI does); otherwise OpenAI's eleven. Gemini's are a fixed list. The voice picked is
+`voice.tts_voice`; podcast hosts use it too (`podcasts.voice_a`, `voice_b`). `POST /voice/speak` takes `{text,
+voice}` to try a voice before keeping it.
+
+`say` only works when Lens runs on the Mac itself (`make start-backend`, not Docker): a container has no `say`. It
+writes WAV, and both podcast hosts get the same voice unless `podcasts.voice_b` names another, since no voice is on
+every Mac.
 
 | Setting | Default |
 | --- | --- |
 | `speech.openai_base_url`, `openai_model` | `https://api.openai.com/v1`, `whisper-1` |
 | `speech.elevenlabs_base_url`, `elevenlabs_model` | `https://api.elevenlabs.io`, `scribe_v1` |
 | `speech.assemblyai_base_url`, `assemblyai_model` | `https://api.assemblyai.com`, `universal` |
 | `speech.deepgram_base_url`, `deepgram_model` | `https://api.deepgram.com`, `nova-3` |
+| `speech.gemini_base_url`, `gemini_model` | `https://generativelanguage.googleapis.com`, `gemini-2.5-flash-preview-tts` |
 | `speech.sentiment` | on |
 | `speech.timeout` | 1800 seconds |
 | `voice.stt` | `same` |
 | `voice.tts_provider` | `openai` |
+| `voice.tts_voice` | the provider's default (alloy, a premade ElevenLabs voice, aura-2-thalia-en, Kore, the Mac's) |
 
 ## What is sent
 

diff --git a/fastapi_backend/app/api/v1/routes/admin.py b/fastapi_backend/app/api/v1/routes/admin.py
@@ -238,7 +238,7 @@ def test_embeddings(user: AdminWriter, cfg: Cfg, db: Db) -> EmbedTestResult:
 
 @router.post("/settings/speech/test")
 def test_speech(
-    user: AdminWriter, cfg: Cfg, provider: str = Query(..., pattern="^(openai|elevenlabs|assemblyai|deepgram)$")
+    user: AdminWriter, cfg: Cfg, provider: str = Query(..., pattern="^(openai|elevenlabs|assemblyai|deepgram|gemini)$")
 ) -> SpeechTestResult:
     """Check a speech provider's address and key with the saved settings (lists its models; nothing is billed)."""
     t0 = time.time()

diff --git a/fastapi_backend/app/api/v1/routes/voice.py b/fastapi_backend/app/api/v1/routes/voice.py
@@ -2,12 +2,12 @@
 
 from __future__ import annotations
 
-from fastapi import APIRouter, HTTPException, Request, Response
+from fastapi import APIRouter, HTTPException, Query, Request, Response
 from starlette.concurrency import run_in_threadpool
 
 from app.api.deps import Cfg, CurrentUser
-from app.domain import voice
-from app.schemas.voice import Heard, SpeakRequest, VoiceInfo
+from app.domain import speech, voice
+from app.schemas.voice import Heard, SpeakRequest, VoiceInfo, VoiceList
 
 router = APIRouter(prefix="/voice", tags=["voice"])
 CLIP = {"requestBody": {"required": True, "content": {"application/octet-stream": {"schema": {"type": "string", "format": "binary"}}}}}
@@ -47,8 +47,19 @@ async def transcribe_clip(request: Request, user: CurrentUser, cfg: Cfg) -> Hear
 async def speak_text(body: SpeakRequest, user: CurrentUser, cfg: Cfg) -> Response:
     """The text read aloud by the text-to-speech model (Settings → AI assistant → Voice). 204 when there's none, or it
     failed: the browser reads it instead."""
-    out = await run_in_threadpool(voice.speak, cfg, body.text)
+    out = await run_in_threadpool(voice.speak, cfg, body.text, body.voice)
     if out is None:
         return Response(status_code=204)
     audio, media = out
     return Response(content=audio, media_type=media, headers={"Cache-Control": "no-store"})
+
+
+@router.get("/voices")
+async def list_voices(
+    user: CurrentUser, cfg: Cfg, provider: str | None = Query(None, pattern="^(" + "|".join(speech.TTS_PROVIDERS) + ")$")
+) -> VoiceList:
+    """The voices a text-to-speech provider (else the one in the voice settings) reads in, from the provider when it
+    lists them (ElevenLabs, Deepgram, macOS say, OpenAI-compatible servers that do), else the ones it's known to have."""
+    p = provider or speech.tts_provider(cfg)
+    found, error = await run_in_threadpool(speech.voices, cfg, p)
+    return VoiceList(provider=p, current=speech.tts_voice(cfg) if p == speech.tts_provider(cfg) else None, voices=found, error=error)
diff --git a/fastapi_backend/app/domain/podcasts.py b/fastapi_backend/app/domain/podcasts.py
@@ -39,7 +39,7 @@
 AFTER = ["analyze", "embed", "summarize"]  # what any new transcript gets
 STATUS = {"gather": "gathering", "plan": "planning", "write": "writing", "check": "checking", "audio": "rendering", "publish": "publishing"}
 # the second host's voice when podcasts.voice_b isn't set, per text-to-speech provider (the first's is voice.tts_voice)
-SECOND_VOICE = {"openai": "onyx", "elevenlabs": "EXAVITQu4vr4xnSDxMKL", "deepgram": "aura-2-orion-en"}
+SECOND_VOICE = {"openai": "onyx", "elevenlabs": "EXAVITQu4vr4xnSDxMKL", "deepgram": "aura-2-orion-en", "gemini": "Puck"}
 PAUSE_MS = {"turn": 450, "same": 250}  # silence before a line: when the other host speaks, and when the same one goes on
 LENGTHS = (5, 10, 20)  # minutes offered; anything from 1 to podcasts.max_minutes goes
 STYLES = {
@@ -718,14 +718,13 @@ def timed(lines, wpm=150, pause_ms=400):
 
 def voices(cfg, req=None):
     """{a, b}: the hosts' voices for the text-to-speech provider: what the episode asked for, else podcasts.voice_a /
-    voice_b, else voice.tts_voice and a second voice of the provider's."""
+    voice_b, else voice.tts_voice and a second voice of the provider's (say has no voice every Mac is sure to have, so
+    both hosts get the same one unless podcasts.voice_b names another)."""
     from . import speech
 
-    v, p = cfg.get("voice") or {}, cfg.get("podcasts") or {}
-    provider = v.get("tts_provider") or "openai"
-    first = {"openai": v.get("tts_voice") or "alloy", "elevenlabs": v.get("tts_voice") or speech.ELEVENLABS_VOICE}.get(
-        provider, v.get("tts_model") or speech.DEEPGRAM_TTS_MODEL
-    )
+    p = cfg.get("podcasts") or {}
+    provider = speech.tts_provider(cfg)
+    first = speech.tts_voice(cfg)  # None for say: the Mac's own voice
     asked = (req or {}).get("voices") or {}
     return {"a": asked.get("a") or p.get("voice_a") or first, "b": asked.get("b") or p.get("voice_b") or SECOND_VOICE.get(provider, first)}
 

diff --git a/fastapi_backend/app/domain/settings.py b/fastapi_backend/app/domain/settings.py
@@ -112,7 +112,7 @@
     "embeddings": ("api_key",),
     "decisions": ("api_key",),
     "voice": ("tts_api_key",),
-    "speech": ("openai_api_key", "elevenlabs_api_key", "assemblyai_api_key", "deepgram_api_key"),
+    "speech": ("openai_api_key", "elevenlabs_api_key", "assemblyai_api_key", "deepgram_api_key", "gemini_api_key"),
     "mail": ("password",),
     "bridge": ("token",),
     "matterbridge": ("slack_token", "discord_token", "telegram_token", "matrix_password"),
@@ -137,7 +137,7 @@
     ("decisions", "engine"): {"auto", "jev", "laya", "llm", "off"},
     ("voice", "input"): {"auto", "server", "browser"},
     ("voice", "stt"): {"same", "sensevoice", "whisper", "mlx-whisper", "openai", "elevenlabs", "assemblyai", "deepgram"},
-    ("voice", "tts_provider"): {"openai", "elevenlabs", "deepgram"},
+    ("voice", "tts_provider"): {"openai", "elevenlabs", "deepgram", "gemini", "say"},
     ("mail", "security"): {"starttls", "ssl", "none"},
     ("bridge", "answer"): {"mention", "all"},
     ("bridge", "approve"): {"off", "low_risk"},