ElevenLabs Speech to Text (2026): How to Transcribe Audio and Video with Scribe v2, Price per Hour, Plans and Limits

ElevenLabs Speech to Text turns audio and video files into text with its Scribe v2 model: 90+ languages, word-level timestamps, speaker labels (diarization) for up to 32 speakers, and tags for sounds like laughter or applause. You can use it two ways: in the ElevenLabs app (upload a file, edit the transcript, export SRT or VTT subtitles), or through the API from your own code. This guide covers both, what each costs, and the limits to check before you upload a long recording.

ElevenLabs Speech to Text at a glance

Open ElevenLabs Speech to Text (PartnerStack referral)

How to transcribe a file in the ElevenLabs app

1. In the ElevenLabs dashboard, open the Speech to Text page and click Transcribe files.
2. Upload an audio or video file.
3. Pick the primary language if you know it, or leave it on Detect so languages in the audio are detected automatically.
4. Turn on Tag audio events if you want laughter, applause and similar sounds marked.
5. Optional: add keyterms (product names, people’s names, jargon) to bias the model toward spelling them right.
6. Click Upload files. When it finishes, click the file name to open the result; click any word to play the audio from that point.
7. Click Export (top right) to download it.

Fix it in the Transcript Editor

Click any transcript to open the Transcript Editor. It is WYSIWYG: click in the text and type. You can also:

• Drag segment handles on the timeline (or type exact times) to adjust start and end timestamps.
• Split a segment with Enter, or merge two adjacent segments from the same speaker.
• Click align words after editing a segment to recompute word-level timestamps.
• Rename speakers, move segments to another speaker one by one or in bulk, add speakers, or delete one (careful: deleting a speaker deletes all of their segments).
• Change playback speed, and add subtitles with the + next to Subtitles in the right-hand panel.

Export formats listed in the docs: Plain text, JSON, HTML, SRT and VTT. SRT and VTT are the subtitle formats most video platforms accept. (The docs screenshot below also shows PDF and DOCX in the menu.)

Need a human check? ElevenLabs’ Productions team offers human-reviewed transcripts from US$2.00 per minute of audio (native-speaker review, optional verbatim mode). The docs describe Productions as beta and available to select users.

How much ElevenLabs Speech to Text costs

ElevenLabs publishes two sets of numbers, depending on where you transcribe.

Through the API: priced per hour of audio

On the API pricing page ElevenLabs says API usage is billed in US dollars, not credits. Speech to Text is billed per audio minute, at these per-hour rates:

The same page lists the hours included with each plan:

Our arithmetic, for scale: a 1-hour podcast costs US$0.22 on Scribe v2; 10 hours cost US$2.20. With keyterms it is US$0.27 an hour, with transcript editing US$0.29 an hour. Prices exclude taxes.

Extras that change the bill (from the docs and API reference):

• Multichannel audio (one speaker per channel, up to 5 channels): each channel is billed at the full audio duration. So a 1-hour, 2-channel file is US$0.44 on Scribe v2 (our arithmetic).
• Transcript editing is billed for at least 10 seconds of audio per request. The guide describes it as a 30% surcharge on the base cost.
• Speaker roles (agent vs customer labels) add 10% to the base transcription cost.
• Entity detection (names, card numbers, medical conditions, 65 entity types) is API-only; the Help Center says it increases the cost by 30%.

In the ElevenLabs app: paid from credits

On the main pricing page, all products share one monthly credit pool, and ElevenLabs gives Speech to Text an approximate cost of 330 credits per minute. If you spent every credit on transcription (our arithmetic, approximate):

On ElevenLabs’ own published numbers, the same Starter plan lists 27 hours of Scribe v2 on the API page but its 30,000 credits cover only about 1.5 hours if you transcribe in the app (our arithmetic). If you transcribe a lot and can run a short script, the API route is listed much cheaper. Check the usage page in your dashboard before you rely on either figure.

See ElevenLabs plans and API pricing

How many files at once (concurrency)

Files longer than 8 minutes are split internally and transcribed in parallel (up to four chunks) to speed up processing.

Features worth knowing

• Keyterm prompting: up to 1,000 keyterms of 50 characters each in batch, or 50 keyterms of 20 characters in realtime. It helps with brand names and jargon.
• No verbatim mode (no_verbatim, Scribe v2): removes filler words, false starts and non-speech sounds. Useful for subtitles and summaries.
• Speaker diarization: up to 32 speakers; you can cap num_speakers if you know how many people talk.
• Transcript editing (experimental): send a plain-English instruction (up to 2,000 characters), for example how to write dates or units, and get edited_transcript back next to the original. It cannot be combined with entity detection, entity redaction or multichannel mode.
• Webhooks: long files can be processed asynchronously, with the result delivered to a webhook you configure.
• Scribe v2 Medical: fine-tuned for clinical audio, with 35% fewer errors there than Scribe v2, at the same price. For HIPAA you must sign a BAA with ElevenLabs Sales first.
• Forced alignment (a separate API): if you already have the text, it returns word timings for it, for example to sync subtitles or an ebook with its audiobook. Same rate as Speech to Text; up to 675,000 characters of text.

Transcribe with the API in a few lines

From ElevenLabs’ quickstart: create an API key in the dashboard, store it as ELEVENLABS_API_KEY, then install the SDK with pip install elevenlabs (there is also a JavaScript SDK and a CLI). The call is:

transcription = elevenlabs.speech_to_text.convert(
    file=audio_data,
    model_id="scribe_v2",
    tag_audio_events=True,
    language_code="eng",   # or None to auto-detect
    diarize=True,
)

For clinical audio, use model_id="scribe_v2_medical"; the request is the same. Realtime transcription uses a WebSocket with scribe_v2_realtime.

Limits, formats and languages

More from ElevenLabs on SiftGuide

For credits and every plan, see the ElevenLabs pricing guide. Want the video in another language rather than a transcript? See our ElevenLabs Dubbing guide. Building a voice assistant that listens and answers live? See the ElevenAgents website guide.

FAQ

Is ElevenLabs Speech to Text free?
Speech to Text is included on the Free plan (US$0, 10,000 credits a month, roughly 30 minutes of audio in the app at about 330 credits per minute, by our arithmetic). The API page lists 4 hours 30 minutes of Scribe v2 for Free / Pay as you go. The Commercial License starts on Starter.

How much does ElevenLabs transcription cost per hour?
Through the API: US$0.22 per hour on Scribe v2 or Scribe v2 Medical, US$0.39 per hour on Scribe v2 Realtime, before taxes. Keyterm prompting and transcript editing cost extra.

Can ElevenLabs transcribe video files?
Yes. Both the app and the API accept video files (MP4, MOV, MKV, WebM and others), up to 3 GB.

Can I get subtitles (SRT or VTT)?
Yes. The Transcript Editor exports SRT and VTT, as well as plain text, JSON and HTML.

How many speakers can Scribe tell apart?
Up to 32 speakers with diarization. You can rename speakers and reassign segments in the editor.

Which languages does it support?
More than 90. Accuracy varies by language; ElevenLabs lists English, Spanish, French, German, Italian, Portuguese, Romanian and many more in its best (“Excellent”) tier.

Is there a human transcription option?
Yes, ElevenLabs Productions offers human-reviewed transcripts from US$2.00 per minute (beta, select users).

Sources

All checked on October 19, 2026. Official ElevenLabs documentation and pricing:

• Transcription (Speech to Text overview, models, limits, languages, FAQ)
• Speech to Text (app guide)
• Transcripts (Transcript Editor)
• Speech to Text quickstart
• Keyterm prompting
• Transcript editing
• Multichannel speech-to-text
• API reference: Create transcript
• Forced Alignment
• Help Center: What is Speech to Text?
• Help Center: How many Speech to Text requests can I make?
• API pricing: https://elevenlabs.io/pricing/api
• Plans and credits: https://elevenlabs.io/pricing