Gladia's real-time API already does something most streaming vendors leave out: it translates on the live socket, and its pricing page says every feature and language is included. So the question is not whether you get translation, but what it costs and what comes with it. On the pay-as-you-go Starter plan, Gladia real-time is $0.75 per hour. The LecSync Realtime Speech API is $0.006 per audio minute, or $0.36 per hour, and adds speaker labels and minutes that build while the session runs.
Everything below comes from public pages read on October 13, 2026: Gladia's pricing page, real-time product page, live session reference, translation docs and concurrency limits. LecSync figures are from the developers page, the pricing docs and the connect docs.
The short answer
Both APIs give you live captions plus a live translation track on one WebSocket. At list price, LecSync costs about half as much: $0.36 per hour against Gladia Starter's $0.75.
The bigger difference is speakers. Gladia's real-time FAQ says real-time diarization isn't supported yet, and suggests sending the recording through the async API afterwards. LecSync puts a speaker label on every final segment during the session. Gladia is the stronger pick if you need more than 60 languages, several target languages in one session, telephony audio, multi-channel input, an SLA or compliance certifications.
How each session works
Both follow the same pattern: your backend makes one authenticated call, then hands the client a socket URL so the API key never ships in the app.
With LecSync, your server calls POST /v1/realtime/connect and gets a stream URL and a one-time token. The client pushes 16 kHz mono PCM. Transcript segments, their translations (paired by segment ID), speaker labels with timestamps (speakers: true) and live minutes (digest: true) come back on that socket.
With Gladia, your server calls POST /v2/live with the audio format and options, and gets back a wss:// URL and a session ID. Translation is turned on under realtime_processing, and each translated utterance arrives as a translation message. A few details shape the integration:
- Translation targets.
target_languagesis an array, so one session can produce several translations. LecSync streams one language pair per session. - Summaries. Gladia's live summarization sits under
post_processingand covers the whole transcription, so it arrives after the session ends. - Speakers. The live init reference has no diarization setting, and the diarization guide is marked pre-recorded only. For calls where each person is on a separate track, Gladia accepts up to 8 audio channels and tags each utterance with its channel.
- Session length. One live session can run for up to 3 hours, then you start a new one.
Side-by-side facts (checked October 13, 2026)
| LecSync Realtime Speech API | Gladia real-time (Starter) | |
|---|---|---|
| Session setup | POST /v1/realtime/connect, then one WebSocket | POST /v2/live, then one WebSocket |
| List price | $0.006/min ($0.36/hr), all features | $0.75/hr, "every feature and language included" |
| Volume price | Same rate | Growth plan from $0.25/hr with an upfront commitment |
| Live languages | 60 | 100+ |
| Live translation | Included; any of 60 into any other, one pair per stream | Included; several target languages per session |
| Speaker labels during the session | Yes, on every final segment, with timestamps | Not supported yet (vendor FAQ); per-channel tags with multi-channel audio |
| Minutes / summary | digest: true, sections pushed during the session | Summary after the session (post_processing) |
| Custom terms | Glossary and domain context, applied to recognition and translation | Phoneme-based custom vocabulary, custom spelling, translation context |
| Latency (vendor figures) | ~0.28 s to first word; translation +~88 ms | Partials under 103 ms; sub-300 ms end to end |
| Audio input | 16 kHz mono PCM (v1) | PCM 8–32 bit, A-law, μ-law; 8–48 kHz; 1–8 channels |
| Billing unit | Accepted audio, fractions of a second | Not stated on the pages we read; multi-channel billed per channel |
| Concurrency | Not published | 30 live sessions on paid plans (1 on free) |
| Max session length | Not published | 3 hours |
| Regions | 6 (dal, hz, jp2, kc, ru2, uk) | 2 (us-west, eu-west) |
| Data retention | Nothing stored unless store: true; then 30 days | Paid default: audio and transcripts kept 3 weeks; zero retention on Enterprise |
| Free allotment | None | €50 one-time credit (about 60 hours of real-time) |
| Compliance (public) | None listed | GDPR, HIPAA, SOC 2 Type 2, ISO 27001 |
| SLA | Best-effort, no SLA today | 99.9% uptime on Growth and Enterprise |
What 1,000 hours cost
All Gladia numbers are list prices. Growth pricing is a "from" figure that depends on your commitment.
| LecSync | Gladia Starter, live only | Gladia Starter, live + async rerun for speakers | Gladia Growth (lowest listed) | |
|---|---|---|---|---|
| Live captions + translation | Included | $0.75/hr | $0.75/hr | $0.25/hr |
| Speaker labels | Included, live | Not available live | +$0.61/hr, after the session | Not available live |
| Minutes during the session | Included | Summary after the session | Summary after the session | Summary after the session |
| Per hour | $0.36 | $0.75 | $1.36 | from $0.25 |
| 1,000 hours | $360 | $750 | $1,360 | from $250 |
Two more things affect the bill. Gladia bills a multi-channel stream at duration times channels, so a two-track call on Starter costs $1.50 per hour. LecSync bills accepted audio down to fractions of a second, and connections that never send audio cost nothing. Gladia's pages don't state a per-session rounding rule, so we left short-stream math out.
For how add-ons stack up across other vendors, see what 1,000 hours cost with translation, speakers and summaries.
Where Gladia is the better pick
- Volume with a commitment. Growth pricing from $0.25 per hour is below LecSync's $0.36.
- Language coverage. 100+ languages with code switching, against LecSync's 60.
- Several translations at once. One session can return captions in, say, English, French and Spanish.
- Phone and multi-track audio. A-law and μ-law, sample rates from 8 to 48 kHz, and up to 8 channels per session.
- Procurement checklists. GDPR, HIPAA, SOC 2 Type 2 and ISO 27001, a 99.9% uptime SLA on Growth, and zero data retention on Enterprise.
- Testing before paying. €50 of free credit and a published concurrency limit of 30 live sessions.
Where LecSync fits better
- Who said what, while it is being said. Every final segment carries a speaker label with timestamps, so a meeting UI can show names live. See building a meeting client with speakers and timestamps.
- Notes that build during the meeting.
digest: truepushes sections of minutes over the same connection; a topic is marked confirmed when it ends. - Lower list price for the full feature set. $0.36 per hour with translation, speakers and minutes, with no plan to negotiate.
- Terms that hold in both languages. The glossary biases recognition and fixes how names and product terms are translated.
- Latency that doesn't drift. After a network stall, the server caps backlog at 2 seconds instead of queueing it. More in why we cap backlog at 2 seconds.
- Wider gateway footprint. Six regions, including Asia, against Gladia's two.
- Nothing kept by default. Storage is off unless you pass
store: true.
LecSync's limits are real: one language pair per stream, 16 kHz mono PCM only in v1, no published concurrency quota or maximum session length, no free allotment, no compliance certifications listed and no SLA today.
FAQ
Does Gladia support real-time translation?
Yes. Set translation: true under realtime_processing with a list of target_languages, and each translated utterance arrives as its own message. Gladia offers base and enhanced translation models, and its pricing page says features are included in the hourly rate.
Does Gladia's real-time API label speakers?
Not yet. Gladia's real-time FAQ says real-time diarization isn't supported and suggests running the recording through the async API afterwards. If each speaker is on a separate audio channel, live utterances carry a channel field instead.
How much does Gladia real-time cost per hour?
$0.75 per hour on the pay-as-you-go Starter plan, checked October 13, 2026. The Growth plan lists real-time from $0.25 per hour with an upfront commitment. New accounts get a one-time €50 credit. Multi-channel streams are billed per channel.
What are Gladia's live limits?
Paid plans allow 30 concurrent live sessions and free accounts allow one. A single live session can last up to 3 hours. Streaming runs in two regions, us-west and eu-west.
What does LecSync include in $0.006 per minute?
Realtime transcription in 60 languages, realtime translation between any two of them, aligned timestamps, speaker labels, glossary and domain context, AI translation polish, live structured minutes, and optional 30-day storage of the recording and transcript. Turning features on or off does not change the rate.
Bottom line
Choose Gladia if you need 100+ languages, several translations per session, telephony or multi-channel audio, compliance paperwork and an SLA, or if you can commit to volume for the Growth rate. Choose LecSync if your product needs captions, a translation track, live speaker labels and running minutes together, at a flat $0.36 per hour. Create a key on the LecSync developers page and open your first stream with the connect docs. Still comparing? See LecSync vs Deepgram and LecSync vs Speechmatics.