Together AI hosts Whisper Large v3 and NVIDIA ASR models for file transcription and realtime WebSocket streaming. Public Transcribe prices on the Together pricing page (checked October 27, 2026) list Whisper Large v3 at $0.0015 per audio minute, Whisper Large v3 (Streaming) at $0.0035 per audio minute, and NVIDIA Nemotron 3 / Parakeet paths from $0.0015 to $0.0045 per minute. Batch jobs for openai/whisper-large-v3 run at up to 50% off serverless rates — Estimated $0.00075 per minute.
The LecSync Realtime Speech API is a managed WebSocket 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 27, 2026 against Together's pricing Transcribe table, the Whisper Large v3 model page ($0.0015/min), transcription overview, streaming, features, translation, batch overview, and credits docs.
What is Together AI Whisper/STT vs a realtime caption API?
Together AI Whisper/STT is a serverless speech recognition stack: HTTP file transcription and translation (POST /v1/audio/transcriptions, /v1/audio/translations), plus a realtime WebSocket at wss://api.together.ai/v1/realtime that returns interim deltas and completed utterances. A realtime caption API such as LecSync keeps one session open for live bilingual captions, speakers and minutes while people are still talking — not only same-language STT text after a file upload or stream.
The short answer
If you need cheap open-model Whisper or NVIDIA ASR — batch files, post-call transcripts, or raw streaming STT — Together undercuts LecSync on dollars per audio minute. Whisper file STT is $0.09/hour Estimated; Streaming Whisper is $0.21/hour Estimated; LecSync is $0.36/hour all-in.
If you need live bilingual captions + speakers + minutes on one flat minute bill while the session runs, LecSync is the closer product. Together's Whisper translate endpoint turns speech into English text (Whisper-style); streaming docs cover interim/final STT, VAD and reconnect — not live any-to-any text MT with speakers and digest on the same socket.
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. Seven gateway regions (dal, jp3, kc, nj, ru2, ru3, uk) keep audio near the capture point; backlog is capped at 2 seconds.
Together AI. File path: multipart or URL upload to /v1/audio/transcriptions (same language) or /v1/audio/translations (speech → English text), models such as openai/whisper-large-v3, Parakeet and Nemotron. Limits: 80 MB direct upload, 1 GB URL fetch, 4 hours max duration per request. File transcriptions support diarize=true with speaker_segments and word-level timestamps. Realtime path: WebSocket streaming with PCM pcm_s16le_16000, interim TranscriptDelta / completed TranscriptCompleted (SDK beta with reconnect and replay). Deepgram Nova-3 / Flux appear as dedicated-only on the models table, not serverless. Platform access requires a prepaid credit balance; docs state no free trial and a minimum $15 first credit purchase.
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 |
| Together Whisper Large v3 | $0.0015 / min | $0.09 | File / serverless Transcribe |
| Together Whisper Large v3 (Streaming) | $0.0035 / min | $0.21 | Realtime WebSocket STT |
| Together Nemotron 3 / Parakeet | $0.0015 / min | $0.09 | Serverless ASR SKUs |
| Together Nemotron 3.5 ASR | $0.0045 / min | $0.27 | Higher-tier ASR list price |
| Together Whisper batch (Estimated) | ~$0.00075 / min | ~$0.045 | 50% off serverless for openai/whisper-large-v3 |
Minute math (Estimated): Whisper → 1000 hours ≈ $90; Streaming → ≈ $210; batch Whisper → ≈ $45. LecSync stays $0.006/min whether translation and speakers are on or off → 1000 hours = $360.
Side-by-side facts (checked October 27, 2026)
| LecSync Realtime Speech API | Together AI Whisper / STT | |
|---|---|---|
| Primary job | Live meeting/lecture captions + MT | Serverless Whisper / NVIDIA ASR (+ streaming STT) |
| Price model | $0.006/min flat, all features | $0.0015–$0.0045/min by model; Streaming Whisper $0.0035 |
| Delivery | Managed WebSocket, preview→final | HTTP file + WebSocket realtime STT |
| Live text translation | Any two of 60, same socket | File /translations → English text; not live any-to-any MT on the socket |
| Speakers | Every final segment (speakers: true) | File diarize=true / speaker_segments; not documented on streaming docs today |
| Live minutes / digest | digest: true | Not available |
| Languages | 60, auto-detected | Whisper-family multilingual; translate task → EN |
| Latency story | ~0.28 s first word; +~88 ms MT | Streaming interim/final STT; file is request/batch |
| Upload / duration limits | Stream while speaking | 80 MB direct / 1 GB URL / 4 h per request |
| Credits | API key billing | Prepaid credits; min $15 first purchase; no free trial (docs) |
| Regions / ops | 7 gateway regions; 2 s backlog cap | Serverless + dedicated endpoints; Deepgram dedicated-only |
What 1,000 hours cost (Estimated)
| Workload, 1,000 hours a month | LecSync | Together AI |
|---|---|---|
| Same-language file / batch transcripts | $360 (live path) | Whisper ~$90; batch Whisper ~$45 |
| Raw streaming STT (no live MT/speakers/digest) | $360 | Streaming Whisper ~$210 |
| Translated live captions (arbitrary pair) | $360 | Not on one Together STT socket; DIY STT + MT |
| Speakers with timestamps during the session | Included | File diarization yes; live WS diarization not documented |
| Live minutes | Included | Not available |
Together wins on raw Whisper and streaming STT dollars, open-model choice and batch cheapness. LecSync wins when the product is a live bilingual caption feed with speakers and minutes, not a serverless ASR job. For flat-vs-add-on math, see what 1,000 hours cost once translation, speakers and summaries are added. For stitching your own STT + MT, see built-in translation vs a DIY pipeline. Nearby Whisper comparisons: LecSync vs Groq Whisper and LecSync vs Cloudflare Workers AI Whisper.
When Together AI is the better pick
- Cheap file or batch Whisper at $0.0015/min serverless, or ~$0.00075/min Estimated on the 50% batch discount for
openai/whisper-large-v3. - Open-model hosting — Whisper, Parakeet, Nemotron serverless; Deepgram Nova-3/Flux on dedicated endpoints.
- Raw WebSocket STT with interim deltas, VAD, and SDK reconnect/replay when you only need same-language text.
- Post-call diarization and word timestamps on the file transcriptions API (
diarize=true). - Large offline jobs within the 80 MB / 1 GB / 4-hour request limits, or chunked URL batches.
When LecSync fits better
- Translated live captions on one socket — any pair of 60 languages, each translation tied to its segment.
- Speakers and recording-aligned timestamps during the live stream — meeting client with speakers and timestamps.
- Live minutes while the session runs (
digest: true). - Predictable minute billing at $0.36 an hour with no separate translation SKU and no waiting for a file round-trip before captions appear.
- Classroom / study-abroad style feeds where captions must keep up as the lecturer switches languages.
FAQ
Is Together Whisper Streaming the same product as LecSync?
No. Together Streaming returns realtime speech-to-text (interim and final). LecSync returns paired transcripts, any-to-any translations, speakers and digest on one connect session at $0.006/min. Treat Together as serverless ASR; treat LecSync as a live caption product.
Does Together streaming include live multilingual text translation?
Public streaming docs cover PCM in, VAD, deltas and completed transcripts. The file /v1/audio/translations path produces English text from speech (Whisper-style). Live captions into an arbitrary target language on the same socket are a LecSync feature, not a documented Together STT add-on.
Does Together WebSocket streaming support speaker diarization?
Speaker diarization is documented on file transcriptions (diarize=true, speaker_segments). The streaming transcription docs reviewed on October 27, 2026 do not document live WebSocket diarization. Do not assume speakers on the realtime path without checking current docs.
Why is LecSync more expensive per hour of audio?
LecSync's $0.36/hour includes live translation, speakers, glossary/polish and digest on the same connection. Together's $0.09–$0.21/hour (Estimated) buys Whisper or Streaming STT. Building live bilingual captions on Together means adding your own MT, live speaker labels and minutes — and owning that glue.
Can I mix both?
Yes. Many teams run LecSync for the live classroom or meeting feed, then push archived audio to Together Whisper (or the batch API) for cheap re-transcription and file diarization.
Next step
Open the LecSync Realtime Speech API docs, create a connect session, and stream three seconds of silence to see preview and final segments on one socket — translation, speakers and digest included at $0.006 per audio minute.
curl -X POST "https://api.lecsync.com/v1/realtime/connect" \
-H "Authorization: Bearer $LECSYNC_API_KEY" \
-H "Content-Type: application/json" \
-d '{"audio":{"encoding":"pcm_s16le","sampleRate":16000,"channels":1},"transcribe":{"languages":["en"]},"store":false}'