Google's Gemini Live API family covers real-time voice agents, a dedicated live transcription model, and a separate speech-to-speech translation SKU. Pricing is token-based (audio at 25 tokens per second) and published as both $/1M tokens and approximate $/min on the Gemini Developer API pricing page.
The LecSync Realtime Speech API is a managed WebSocket built for meeting and lecture captions: preview and final text, translation between any two of 60 languages, a speaker and audio-aligned timestamps on every final segment, and running minutes. It costs $0.006 per audio minute ($0.36 per hour) with those features included.
Figures below were checked on October 24, 2026, against Google's Live API overview, Live Transcription guide, capabilities, session management, best practices, and the pricing page (last updated October 22, 2026 UTC).
What is Gemini Live API vs a realtime STT caption API?
Gemini Live API is Google's bidirectional WebSocket for low-latency audio (and optional video) sessions. The same transport powers conversational voice agents such as gemini-3.8-live, a dedicated speech-to-text path gemini-3.5-transcribe-live, and speech-to-speech translation via gemini-3.5-live-translate-preview. A realtime STT caption API such as LecSync streams one audio input and returns paired transcripts, translations, speakers and minutes without running a conversational agent.
The short answer
If you are building a voice agent that listens, reasons, speaks back, calls tools, or sits inside the Google AI Studio / Gemini ecosystem, Gemini Live is the natural fit. gemini-3.8-live lists about $0.005 per minute of audio input and $0.018 per minute of audio output, with barge-in, proactive audio, and multimodal input.
If you need live translated captions, speakers and minutes for meetings or lectures on one flat minute bill, LecSync is the closer product. Same-language live STT on Gemini's gemini-3.5-transcribe-live is estimated by Google at about $0.009 per minute — already above LecSync's $0.006 all-in — and that SKU does not add live translation or live speaker labels. Speech-to-speech Live Translate is a different model at roughly $0.037 per minute.
How each API works
LecSync. Your server calls POST https://api.lecsync.com/v1/realtime/connect with the audio format, a language pair and the features you want, then streams 16 kHz mono PCM. Preview text arrives while someone is still talking; a final segment settles once the sentence confirms. Translations share the segment ID. With speakers: true, each final carries a speaker and startMs/endMs on the recording clock. With digest: true, minutes arrive section by section. The first word shows up about 0.28 seconds after speech starts; translation adds roughly 88 ms.
Gemini Live Agent (gemini-3.8-live). You open a Live WebSocket (Gen AI SDK or raw WSS), set response_modalities to AUDIO, and stream PCM in. The model replies with spoken audio (24 kHz PCM), optional input/output text transcriptions, tool calls and Google Search. Sessions without context compression stop at 15 minutes for audio-only (2 minutes with video); connections reset around 10 minutes, so production apps enable compression and session resumption.
Gemini Live Transcription (gemini-3.5-transcribe-live). Same Live transport, but configured as a speech-to-text pipeline with response_modalities=["TEXT"]. You get interim_input_transcription while someone is speaking and input_transcription when a turn finalizes. Auto language detection covers 85+ languages with code-switching; custom vocabulary and SMART formatting are available. Continuous streaming is documented up to 10 minutes per session. Speaker diarization and word-level timestamps are not supported on the live path — Google points you to the non-streaming gemini-3.5-transcribe endpoint for those.
Pricing converted to dollars per hour (Estimated)
| Path | Published rate | Implied $/hour of audio | Notes |
|---|---|---|---|
| LecSync Realtime Speech API | $0.006 / min all-in | $0.36 | STT + translation + speakers + digest included |
| Gemini 3.5 Transcribe Live | ~$0.009 / min blended (Google estimate) | ~$0.54 | STT only; 25 audio tokens/s + ~175 text tokens/min |
| Gemini 3.8 Live, audio in only | $0.005 / min list | $0.30 | Voice-agent listening list price; compounding can raise real spend |
| Gemini 3.8 Live, audio in + out | $0.005 + $0.018 / min | $1.38 | Assumes continuous reply audio; agent product |
| Gemini 3.5 Live Translate | ~$0.0368 / min blended | ~$2.21 | Speech-to-speech, 70+ languages |
Agent sessions re-bill accumulated context each turn, and enabling input/output transcription adds text-output charges on top of audio tokens. Treat the agent rows as floors, not meeting-caption quotes.
Side-by-side facts (checked October 24, 2026)
| LecSync Realtime Speech API | Gemini Live (Agent / Transcribe / Translate) | |
|---|---|---|
| Primary job | Meeting/lecture captions + MT | Voice agents; dedicated Live STT; S2S translate |
| Price model | $0.006/min flat, all features | Token $/1M; Google publishes ~$/min estimates |
| Live STT (same-language) | Included | gemini-3.5-transcribe-live ~$0.009/min |
| Live text translation | Any two of 60, same socket | Not on Transcribe Live; S2S via Live Translate ~$0.037/min |
| Speakers on live stream | Every final segment | Live Transcribe: no; file Transcribe: ≤8 |
| Live minutes / digest | digest: true | No equivalent on Live Transcribe |
| Interim results | Preview → final | interim + final transcription fields |
| Languages | 60, auto-detected | Live Transcribe 85+; Live Agent docs cite 70 / 99-table |
| Session length | Long-running managed streams; 2 s backlog cap | Live STT ≤10 min; agent 15 min unless compression |
| Hosting | Managed SaaS, 7 gateway regions | Google-hosted Developer API (Preview) |
| Custom vocabulary | Glossary + domain context | Up to 1,000 terms (best ≤100) |
| Audio format | 16 kHz mono PCM (typical) | 16-bit PCM 16 kHz in; agent out 24 kHz |
| Uptime commitment | Best effort, no SLA | Preview; enterprise options via Google Cloud paths |
What 1,000 hours cost
| Workload, 1,000 hours a month | LecSync | Gemini (Estimated from public list) |
|---|---|---|
| Same-language live captions | $360 | Transcribe Live ~$540 |
| Translated live captions (text) | $360 | Not on Transcribe Live; Live Translate S2S ~$2,208 or DIY STT+MT |
| Speakers with timestamps | Included | Live: not available; file Transcribe only |
| Live minutes | Included | Not available |
| Voice agent (listen + talk back) | Not the product | 3.8 Live from ~$0.30–$1.38+/hr depending on output |
1,000 hours is 60,000 minutes. LecSync stays $360 whether you turn translation and speakers on or off. Gemini's closest live caption SKU is already about 50% higher before you add translation or speakers. For a deeper flat-vs-add-on walkthrough, see what 1,000 hours cost once translation, speakers and summaries are added.
When Gemini Live is the better pick
- Voice agents and multimodal dialogue. Barge-in, affective dialog, proactive audio, tools and Google Search on
gemini-3.8-live. - Speech-to-speech translation.
gemini-3.5-live-translate-previewis built for spoken reply audio across 70+ languages. - Google ecosystem. AI Studio, Gen AI SDK, ephemeral tokens, and partner stacks (LiveKit, Pipecat, Agora and others listed in the Live Transcribe guide).
- Smart / verbatim formatting and large vocabulary lists on the dedicated Transcribe Live model.
- Broader language coverage on STT alone (85+ on Transcribe Live vs LecSync's 60 with built-in MT).
When LecSync fits better
- Translated live captions on one socket. Any pair of 60 languages, each translation tied to its segment — see built-in translation vs a DIY STT + MT pipeline.
- Speakers and recording-aligned timestamps during the live stream, not only on a post-call file job — meeting client with speakers and timestamps.
- Live minutes while the session runs (
digest: true) — live digest vs post-call summarization. - Predictable minute billing at $0.36 an hour with no compounding context window and no separate translation SKU.
- Long lectures without 10-minute live-STT caps, with a 2-second backlog cap instead of dropping behind.
Comparing other Google-adjacent or realtime stacks? See LecSync vs Google Cloud Speech streaming and LecSync vs AssemblyAI / OpenAI Realtime on one connection.
FAQ
How much does Gemini Live API cost for audio?
As of October 24, 2026, gemini-3.8-live lists $3.00 per 1M audio input tokens ($0.005/min) and $12.00 per 1M audio output tokens ($0.018/min). gemini-3.5-transcribe-live is estimated by Google at about $0.009 per minute blended. Live Translate is about $0.037 per minute. Real agent spend can rise with context re-billing and transcription surcharges.
Does Gemini Live Transcribe translate captions in real time?
No. gemini-3.5-transcribe-live returns speech-to-text. Live Translation is a separate speech-to-speech model (gemini-3.5-live-translate-preview). LecSync returns any-to-any text translation of 60 languages on the same caption stream.
Can Gemini label speakers on a live stream?
Not on Live Transcription. Google documents speaker diarization (up to eight speakers) on the non-streaming gemini-3.5-transcribe path only. LecSync attaches a speaker and millisecond timestamps to every live final segment.
How long can a Gemini Live session run?
Without compression, audio-only agent sessions are limited to 15 minutes (2 minutes with video) and connections reset around 10 minutes. Live Transcription documents continuous streaming up to 10 minutes. Context window compression plus session resumption extends agent sessions; LecSync is designed for long managed caption streams with a 2-second backlog cap.
Bottom line
Choose Gemini Live when you need a conversational voice agent, speech-to-speech translation, or deep Google AI tooling. Choose LecSync when you need translated captions, speakers and live minutes on one flat $0.006/min bill without assembling multiple Gemini SKUs. You can open a test stream from the developers page with one POST /v1/realtime/connect.