Skip to content

feat(livekit-voice-agent): update to Lightning v3.1 Pro #52

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
harshitajain165 merged 2 commits into from
Jun 1, 2026
Merged

feat(livekit-voice-agent): update to Lightning v3.1 Pro #52

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b3dfbab
feat(livekit-voice-agent): update to Lightning v3.1 Pro as default TT…
harshitajain165 May 21, 2026
dc64ceb
feat(livekit-voice-agent): remove StreamAdapter — TTS now streams nat…
harshitajain165 Jun 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 7 additions & 10 deletions voice-agents/livekit-voice-agent/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,7 +81,7 @@ Open the [LiveKit Agents Playground](https://agents-playground.livekit.io) and e

| Variable | Default | Description |
|---|---|---|
| `VOICE_ID` | `sophia` | TTS voice ID — browse all voices at [waves.smallest.ai](https://waves.smallest.ai) |
| `VOICE_ID` | `meher` | TTS voice ID — browse all voices at [waves.smallest.ai](https://waves.smallest.ai) |
| `LANGUAGE` | `en` | BCP-47 language code for STT and TTS. Use `multi` for automatic detection across 39 languages |
| `LLM_MODEL` | `gpt-4o-mini` | OpenAI model name |

Expand All @@ -97,18 +97,15 @@ Real-time transcription over WebSocket with ~64ms TTFT. Set `language="multi"` t
stt=smallestai.STT(language="en")
```

### Smallest AI TTS — Lightning
### Smallest AI TTS — Lightning v3.1 Pro

The Lightning TTS plugin synthesizes audio per sentence rather than streaming tokens, so it is wrapped in `tts.StreamAdapter` with a `SentenceTokenizer`. The adapter buffers LLM output to the next sentence boundary before firing synthesis — keeping first-audio latency low while the LLM is still generating:
The plugin uses persistent WebSocket streaming backed by a connection pool for low-latency audio delivery — no `StreamAdapter` wrapper needed:

```python
tts=tts.StreamAdapter(
tts=smallestai.TTS(
model="lightning-v3.1",
voice_id="sophia",
language="en",
),
sentence_tokenizer=tokenize.basic.SentenceTokenizer(),
tts=smallestai.TTS(
model="lightning_v3.1_pro",
voice_id="meher",
language="en",
)
```

Expand Down
15 changes: 5 additions & 10 deletions voice-agents/livekit-voice-agent/agent.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,15 +12,13 @@
WorkerOptions,
cli,
metrics,
tts,
tokenize,
)
from livekit.plugins import openai, silero, smallestai

logger = logging.getLogger("livekit-voice-agent")
load_dotenv()

VOICE_ID = os.getenv("VOICE_ID", "sophia")
VOICE_ID = os.getenv("VOICE_ID", "meher")
LANGUAGE = os.getenv("LANGUAGE", "en")
LLM_MODEL = os.getenv("LLM_MODEL", "gpt-4o-mini")

Expand Down Expand Up @@ -50,13 +48,10 @@ async def entrypoint(ctx: JobContext):
vad=ctx.proc.userdata["vad"],
stt=smallestai.STT(language=LANGUAGE),
llm=openai.LLM(model=LLM_MODEL),
tts=tts.StreamAdapter(
tts=smallestai.TTS(
model="lightning-v3.1",
voice_id=VOICE_ID,
language=LANGUAGE,
),
sentence_tokenizer=tokenize.basic.SentenceTokenizer(),
tts=smallestai.TTS(
model="lightning_v3.1_pro",
voice_id=VOICE_ID,
language=LANGUAGE,
),
)

Expand Down