Speech to Text

What it does

Speech to Text takes uploaded audio or video files, or a live audio stream, and turns the speech into text. Batch transcription returns a transcript with word-level timestamps, speaker labels, and audio event tags such as laughter. Realtime transcription, using Scribe v2 Realtime, returns text while speech is still being spoken, which suits live calls and voice agents.

People use the Studio interface to upload recordings and export captions, subtitles, or editable transcripts for podcasts, interviews, and video. Developers call the API or SDKs to build transcription into their own products. Optional features include keyterm prompting for product names and specialized terms, entity detection for items such as credit card numbers, speaker diarization for up to 32 speakers, and transcript editing through natural-language instructions.

Suits teams that produce recorded or live speech content and need searchable or captioned output in many languages. Access requires an ElevenLabs account; the free tier draws on the shared credit pool and paid plans start at $6 per month. Limits include a 3 GB file size cap, a 10-hour standard duration limit, and billing for silence. Keyterm prompting and transcript editing carry extra charges, and HIPAA use requires a Business Associate Agreement through sales.

What it can do

• Batch transcription of uploaded audio and video files with word-level timestamps
• Realtime transcription with low latency (about 150 ms) via streaming
• Speaker diarization for up to 32 speakers
• Keyterm prompting for up to 1000 terms in batch mode
• Entity detection across 65 entity types with timestamps
• Transcript editing via natural-language instructions (experimental)

Works with

• ElevenLabs Agents for real-time voice interactions
• ElevenLabs Studio for editable text and captions
• Webhooks for asynchronous transcription delivery

Example workflows

Caption a recorded video

Input: an MP4 or MOV file. Steps: upload it in Studio, let Scribe v2 process the audio, review the transcript with timestamps, then export captions or subtitles. Output: an editable transcript with captions.

Live call transcription for an agent

Input: a live audio stream from a phone or web call. Steps: connect the stream to Scribe v2 Realtime over WebSockets, consume partial and committed text as speech arrives, and pass the text to a downstream agent. Output: text delivered in near real time.

Before you choose

• Silence counts toward billed duration, and keyterm prompting and transcript editing carry extra costs.
• HIPAA-regulated use requires a Business Associate Agreement signed through ElevenLabs sales before deployment.
• Language accuracy varies by language; many languages fall in moderate or good accuracy bands rather than excellent.

Listed under

• Productivity
• Meetings and notes
• Captions
• Developers
• Media
• Real-time
• Real-time transcription
• Speech to text

The free plan includes 10,000 credits per month. Paid plans are monthly or yearly with yearly billed at ten months' price. Credits are shared across products. The API speech-to-text rate is usage-based; FAQ text cites from $0.40 per hour and the API pricing page holds the rate table.

Source: elevenlabs.io/pricing, checked 28 Oct 2026. Prices can change; check before you buy.