AIDealsPro is reader-supported. Some links on this page are affiliate links: if you buy through them we may earn a commission at no extra cost to you. It never changes our verdicts. How we review
ElevenLabs speech to text runs on Scribe, the platform’s transcription model: upload audio or send it via the API and get back text with word-level timestamps. Billing is credit-based like the rest of ElevenLabs, so a 10-minute file costs a fraction of the free tier’s 10,000 monthly credits. Checked 4 Nov 2026.
Best for: Turning interviews, meetings, and video audio into accurate, timestamped transcripts without leaving ElevenLabs.
Not for: Anyone expecting transcription to be free — it spends credits like everything else on the platform.
Starting price: $0 — Scribe works on the free 10,000-credit tier; paid plans from $6/month, checked 4 Nov 2026.
What is ElevenLabs speech to text?
ElevenLabs speech to text is Scribe, the company’s transcription model. Where the text-to-speech side turns writing into audio, Scribe runs the other direction: it takes an audio file — or an audio stream through the API — and returns written text, complete with word-level timestamps that mark when each word was spoken.
The timestamps are what make it more than a dictation tool. Word-level timing lets you build captions, clickable chapters, and searchable archives from a single transcription pass, instead of paying separately for transcription and then for timing.
Scribe handles dozens of languages, which matters for interviews and content recorded outside English. Language detection is automatic on upload, so there is no settings page to configure before your first file — pick the audio and go.
How to transcribe audio in the ElevenLabs app in 5 steps
The in-app route needs no code. Credit costs are noted per step:
- Open the speech-to-text section of the ElevenLabs app and choose the Scribe model. Cost: 0 credits.
- Upload your audio file (or point Scribe at a recording). Cost: 0 credits — nothing is billed until transcription runs.
- Confirm the language, or leave automatic detection switched on. Cost: 0 credits.
- Run the transcription. Cost: credits per audio minute, billed from your monthly pool — check the current per-minute rate in the app before a large batch.
- Review the transcript with its word-level timestamps, then export it as text or captions. Cost: 0 credits for review and export.
Only step 4 spends credits. Uploading, language detection, reviewing, and exporting are all free — the meter runs solely on the minutes Scribe actually transcribes.
How to get the best transcription quality (4 rules)
Scribe is only as good as the audio you feed it. These four rules lift accuracy on any transcription tool:
- Record clean audio. A close microphone in a quiet room beats any post-processing — background music and room echo are the top accuracy killers.
- Separate speakers when you can. One voice per recording transcribes far better than crosstalk; for interviews, use individual tracks if your recorder supports them.
- Avoid heavy compression. Low-bitrate audio strips the detail transcription models rely on — export at the highest quality your workflow allows.
- Keep a natural pace. Rushed, overlapping speech is the hardest input for any model, including Scribe — ask guests to slow down rather than talk over each other.
If a transcript comes back messy, fix the audio before blaming the model. Re-record the worst five minutes cleanly and compare — the difference tells you whether the problem was the input or the tool.
For recurring shows, turn these rules into a recording checklist and share it with every guest. Consistent input quality is what turns transcription from a chore into a one-pass task.
How much does ElevenLabs speech to text cost?
Transcription draws from the same monthly credit pool as text-to-speech: the free tier’s 10,000 credits, Starter’s 30,000, Creator’s 121,000, or Pro’s 600,000. There is no separate transcription subscription to buy — one balance covers both directions.
The billing unit differs, though. Text-to-speech counts input characters (one credit per character); transcription counts audio minutes at a per-minute credit rate shown in the app and API docs. Because the two rates are separate, always check the current per-minute transcription rate before estimating a big batch — do not assume the 1-credit-per-character TTS math applies.
For planning, the arithmetic is simple once you have the rate: minutes of audio multiplied by credits per minute equals the job’s cost. A 60-minute interview at a hypothetical 100 credits per minute would cost 6,000 credits — 60% of a free-tier month, or 20% of Starter’s 30,000. Plug in the real rate from the app and the same formula holds.
Scribe vs text-to-speech: how the credit model differs
Think of it as two meters on one balance. The TTS meter counts characters going in; the transcription meter counts audio minutes going in. Both deduct from the same 10,000 / 30,000 / 121,000 / 600,000 monthly allowance, and both reset monthly with no rollover.
| Direction | Billing unit | Example |
|---|---|---|
| Text to speech | 1 credit per character | 5,000-character script = 5,000 credits |
| Speech to text | Credits per audio minute | 60-minute file = 60 × per-minute rate |
The practical consequence: a mixed workflow is easy to budget. A podcaster who transcribes a 60-minute episode and then generates a 2-minute promo from the transcript can estimate both legs from the same credit balance, with no second subscription involved. Our ElevenLabs API pricing guide covers the same math for automated pipelines.
What can you do with the transcript?
Captions are the most common output. The word-level timestamps map directly onto subtitle files, so a transcribed video becomes an accessible, captioned video with one export — no manual timing pass.
Searchable archives are the second win. A transcribed interview library becomes text you can search, quote, and repurpose: pull highlights for social clips, feed quotes into articles, or build show notes without re-listening to hours of audio.
The third use is the round trip: transcribe source audio, edit the text, then generate new voiceover from the edited script with ElevenLabs text-to-speech. Interview to article to narrated video becomes a single-platform workflow on one credit balance.
ElevenLabs speech to text: pros and cons
Pros
- Word-level timestamps included — captions and chapters come from the same pass.
- Dozens of languages with automatic detection; no configuration before the first file.
- Shares one credit pool with text-to-speech — no second subscription.
- Works on the free tier, so testing transcription costs nothing extra.
Cons
- Not free: transcription spends credits at its own per-minute rate.
- Accuracy drops with background noise, overlapping speakers, and heavy accents — like any transcription tool.
- The per-minute rate is separate from the familiar 1-credit-per-character TTS rate, which confuses budgeting.
- Very long files need to fit inside the monthly credit pool — an hour of audio is a real chunk of the free tier.
Who should NOT use ElevenLabs for transcription
If you never generate speech, a transcription-only workflow may be simpler elsewhere — you would be paying for a voice platform to use half of it. ElevenLabs earns its place when transcription feeds generation: captions plus voiceover, interviews plus narrated cuts.
And if you transcribe hundreds of hours a month, check the per-minute rate against your plan before committing. At that volume the credit math decides everything — see our ElevenLabs review for how the plans compare at scale.
Frequently asked questions
Does ElevenLabs do speech to text?
Yes. ElevenLabs speech to text runs on Scribe, its transcription model, available in the app and via the API on every plan. Upload audio and get back text with word-level timestamps, billed in credits from the same monthly pool as text-to-speech.
How accurate is ElevenLabs Scribe?
Accuracy depends on the audio. Clean, single-speaker recordings with little background noise transcribe best; overlapping speakers, heavy accents, and noisy environments reduce quality on Scribe exactly as they do on any transcription tool. Test with your own typical audio before a big batch.
How much does ElevenLabs transcription cost?
It costs credits per audio minute, drawn from your monthly pool — 10,000 on the free tier, 30,000 on Starter ($6/month), and up. The per-minute rate is listed in the app and API docs and differs from the 1-credit-per-character TTS rate, so check it before estimating.
Can I transcribe YouTube videos with ElevenLabs?
Yes, if you have the audio file — download or extract the audio first, then upload it to Scribe. Only transcribe content you have the rights to; the tool does not check ownership, so the responsibility for rights sits with you.
Does ElevenLabs speech to text support multiple languages?
Yes. Scribe handles dozens of languages with automatic language detection, so there is nothing to configure before uploading. Mixed-language audio is the hardest case — expect lower accuracy where speakers switch languages mid-sentence compared to single-language recordings, and plan a review pass for mixed audio.
Can I get timestamps with my ElevenLabs transcript?
Yes — word-level timestamps are built into Scribe’s output, not an add-on. They mark when each word was spoken, which is what makes caption files, clickable chapters, and searchable archives possible from a single transcription without any extra step or extra cost.
Sources checked
Scribe features, word-level timestamps, and language support come from the official ElevenLabs website and documentation, checked 4 Nov 2026. Plan prices and credit allowances come from the official ElevenLabs pricing page, checked 4 Nov 2026.
Last updated 4 Nov 2026. Prices checked on the official ElevenLabs pricing page on that date; confirm at checkout.