Picking a Soniox alternative realtime speech API is rarely about “who can transcribe.” Both vendors ship realtime streaming STT with translation in the product story. The harder questions are how the bill is calculated, and how many concurrent live sessions you can open before the API says no.
Quick verdict: Soniox is a credible realtime speech stack—streaming STT, translation in the same call, diarization and formatting bundled, with public list rates that market around $0.12/hour of realtime transcription. Choose LecSync Realtime STT when you want a flat $0.006 per audio minute (all listed features included) and a concurrency model aimed at shipping a product, not a 10-stream default ceiling.
What is a realtime speech API with predictable minute pricing?
A realtime speech API with predictable minute pricing charges for accepted audio duration at a published per-minute (or per-hour) rate you can multiply by session length—without reconstructing cost from separate input-audio, input-text, and output-text token meters. LecSync publishes one all-in rate on its developers page; Soniox’s public pricing states that all API costs are calculated based on tokens, with hourly equivalents shown for orientation.
Soniox strengths (worth acknowledging first)
Before contrasting billing and limits, credit what Soniox’s public materials emphasize:
- Realtime + async STT across 60+ languages, with a pricing calculator that puts realtime STT near $0.12/hour and async near $0.10/hour (pay-as-you-go assumptions; updated June 2026 on their pricing page).
- Translation in the same realtime API call at no separate translation line item, per their FAQ—useful if you want captions without wiring a second MT vendor.
- Diarization, language identification, and smart formatting described as bundled into the hourly rate rather than sold as add-ons.
- A full speech stack narrative (STT plus TTS on the same site), with docs for concurrency monitoring and limit increases in the Console.
If you are evaluating quality demos, multilingual streaming, or a low advertised hourly STT rate, Soniox belongs on the shortlist. The rest of this article is about production economics and default quotas—not a claim that Soniox “cannot transcribe.”
Side-by-side: LecSync vs Soniox realtime STT
Figures below were rechecked on public pages ~2026-10-01 (Asia/Shanghai). They are vendor-published list facts, not an independent bake-off.
| Dimension | LecSync Realtime STT | Soniox (public pricing / STT realtime docs) |
|---|---|---|
| Connection model | One WebSocket: transcription + translation paired server-side | Realtime WebSocket STT; translation described as same realtime API call |
| Languages | 60, any-to-any translation pairs | 60+ languages (pricing page) |
| First word on screen | 0.28 s after speech starts (public product figure); translation ~+88 ms to first translation token | Not published as a single comparable “first word” number on the pricing page |
| Speakers / diarization | Labels on final segments when speakers enabled, with timestamps | Speaker diarization bundled (pricing FAQ) |
| Digest / minutes | digest:true → structured minutes on the same connection | Not the same “digest on same WS” product surface on the public pricing page |
| Backlog under hitch | Cap 2 s; captions catch the current sentence within ~1 s after recovery | Not described the same way on the public pricing page |
| Regions | 5 gateways: dal, hz, jp2, ru, uk | Region-scoped concurrency APIs documented |
| Pricing shape | Flat $0.006 / audio minute, listed features included | Token-based; FAQ/calculator equivalent ~$0.12/hour realtime STT |
| Default realtime concurrency | Product API positioned for multi-session apps (see developers) | Concurrent requests: 10 simultaneous WebSocket connections (STT Limits & quotas) |
Sources: LecSync developers, Soniox pricing, Soniox STT Limits & quotas.
Pricing: tokens vs a flat audio minute
Soniox’s pricing page is explicit: “All API costs are calculated based on tokens.” Realtime (streaming) rates shown publicly include:
- Input audio tokens: $2.00 per 1M tokens
- Input text tokens (custom instructions / context): $4.00 per 1M tokens
- Output text tokens (transcription and optionally translation): $4.00 per 1M tokens
They also publish usage references (about 30,000 input audio tokens per hour of audio; about 15,000 output text tokens per hour of speech) and market an equivalent of roughly $0.12/hour for realtime transcription. That hourly figure is helpful for orientation—but the meter is still token-shaped. Custom context, longer transcripts, or translation output can move the bill away from a simple “minutes × rate” spreadsheet. Advanced cases with translation or custom context are called out as billed by token usage.
LecSync’s public developers page states a single price: $0.006 per minute of audio, with realtime transcription, realtime translation, audio-aligned timestamps, speaker separation, glossary, AI translation polish, live structured summary, and optional 30-day recording/transcript storage included. Disabling features does not lower the rate; enabling all of them does not raise it. Fractional seconds are charged proportionally; no accepted audio means no charge.
Worked example (public list orientation)
| Stack | How the public page frames cost | 60 minutes of audio (orientation) |
|---|---|---|
| Soniox realtime STT (hourly equivalent on pricing / FAQ) | ~$0.12/hour marketed equivalent; actual invoice is token-based | ~$0.12 if you land on that equivalent—plus variance if token mix differs |
| LecSync all-in | $0.006 / audio minute flat | $0.36 |
Soniox’s marketed hourly STT rate can look cheaper on a pure-transcription spreadsheet. The product question is whether you can forecast that number when output tokens, translation text, and custom instructions also meter—and whether a default of 10 concurrent sockets is enough when your app opens many rooms at once.
Concurrency: default 10 simultaneous WebSockets
Soniox’s realtime STT Limits & quotas document lists:
| Limit | Value | Notes |
|---|---|---|
| Requests per minute | 100 | Exceeding may rate-limit |
| Concurrent requests | 10 | Maximum simultaneous active WebSocket connections |
| Stream duration | 300 minutes | Fixed; cannot be increased—open a new session to continue |
The same page says you can request higher limits (except stream duration) in the Soniox Console. That path exists—but the default concurrent cap is 10. For a production product that hosts many simultaneous live caption or meeting sessions, ten open WebSockets is a low starting threshold: a classroom product, a webinar platform, or a multi-tenant meeting client can hit it during ordinary peaks before you finish sales conversations about limit raises.
LecSync’s public developers materials position the Realtime Speech API for product integrations (multi-region gateways, backlog capping, digest on the same connection) rather than a documented default of ten concurrent sockets. When concurrency is the gating risk, verify current limits with each vendor before launch—and treat Soniox’s published default of 10 as a hard planning input until a higher quota is confirmed in writing.
When to pick Soniox
Pick Soniox when:
- You want a vendor whose public story centers on efficient realtime STT with translation and diarization described as included
- Your concurrency needs fit inside 10 simultaneous WebSocket sessions (or you already have a Console increase approved)
- You are comfortable modeling cost from token tables (audio in, text in, text out) rather than a single audio-minute line
- You may also use their TTS surface on the same account
A prototype, a single-room demo, or a low-concurrency internal tool can sit comfortably inside those defaults.
When to pick LecSync
Pick LecSync when the brief says ship live bilingual (or any-to-any) captions as a product:
- Classroom, webinar, or meeting clients that show source + target lines as people speak
- Apps that need speaker-labelled finals and structured
digestminutes without a second summarization vendor - Teams that want one invoice line at $0.006 / audio minute instead of reconstructing spend from token mix
- Launches where default 10-stream concurrency would be a production incident waiting to happen
Connect with POST https://api.lecsync.com/v1/realtime/connect and a Bearer key, then stream on the returned WebSocket (developers). Preview text streams, then finals; language auto-detect is available.
Integration scenarios
Multi-room live captions. Ten concurrent Soniox WebSockets may be fine for a pilot. A campus product or SaaS meeting tier that opens many rooms in the same minute needs a higher confirmed quota—or an API whose defaults match that shape.
Forecastable COGS. Finance asks “what does 10,000 caption-minutes cost next quarter?” LecSync’s answer is multiplication: minutes × $0.006. Soniox’s answer depends on token mix even when the marketing equivalent is ~$0.12/hour.
Translation on the same pipe. Both vendors describe translation alongside realtime STT. LecSync documents paired transcription + translation on one WebSocket with a flat minute price; Soniox describes translation in the same realtime call with token-based billing for advanced usage.
FAQ
Is Soniox cheaper than LecSync for realtime STT?
On Soniox’s marketed ~$0.12/hour realtime equivalent, pure STT can look lower than LecSync’s $0.36 per 60 audio minutes. LecSync’s pitch is predictability and an all-in feature set at $0.006/minute, not winning every raw-STT spreadsheet row. Token mix (context, output length, translation text) can move Soniox’s invoice away from the hourly equivalent.
Does Soniox bill by the minute or by the token?
Public pricing says all API costs are calculated based on tokens, with hourly equivalents (~$0.10 async / ~$0.12 realtime) for orientation. Realtime token rows include input audio, input text, and output text at per-1M rates.
What is Soniox’s default realtime concurrency limit?
The STT realtime Limits & quotas page lists concurrent requests: 10—the maximum number of simultaneous active WebSocket connections. Higher limits (except the 300-minute stream duration) can be requested in the Console.
How much is LecSync Realtime STT?
$0.006 per audio minute, all listed features included, on the public developers page. No accepted audio means no charge.
How fast is LecSync’s first caption word?
LecSync publishes 0.28 seconds from speech start to first word on screen, and about 88 ms added to the first translation token. Treat those as vendor product figures, not a lab bake-off against Soniox.
Does turning features off make LecSync cheaper?
No. The public rate stays $0.006 per audio minute. Disabling translation, speakers, or digest does not reduce it; enabling all listed features does not increase it.
Verdict
Soniox earns a fair look for realtime multilingual STT with translation and diarization in the product story, and for a low advertised hourly STT equivalent—if your concurrency fits the default of 10 WebSockets (or a raised quota) and you accept token-based billing. LecSync fits when you need live translation, speakers, and digest on one WebSocket at a flat $0.006 per audio minute, with concurrency expectations closer to a shipped product than a ten-stream ceiling. Start at the LecSync Realtime STT docs and compare line items against Soniox pricing and Limits & quotas for your peak session count.