ElevenLabs Scribe: Turn Recordings Into Text and Keep Going
Drop a podcast episode, an interview or a meeting recording into the chat, and ElevenLabs Scribe turns it into text. With Scribe v2, the transcript also shows who's talking and when each word was said.
On Cabina.AI, ElevenLabs speech-to-text works in one chat with Claude, Gemini and other top models for text, images, audio and video. That means the transcript isn't where the job ends. You can hand it to another model in the same conversation and keep working on it right there.
You don't have to figure out ElevenLabs Scribe speech-to-text pricing before you've seen a single result. Start on the Free plan at $0.00 per month and run Scribe on your own recording first. Upgrade later, once you know it fits your work.
Transcribe Your Recording Free on Cabina.AIWhat You Can Do With Scribe
Scribe v2 is the current model behind the ElevenLabs Scribe speech-to-text product, and ElevenLabs built these features into it.
Speaker Labels for Up to 32 People
Speaker labels show who said what, for up to 32 speakers in one file. It's one of the ElevenLabs speech-to-text features that make meetings, interviews and podcasts with many voices easy to follow.
Word-Level Timestamps
Each ElevenLabs Scribe transcription can mark the time of every word, and the model also supports character-level timing. You'll find a single quote in a long recording without scrubbing through it.
Audio Event Tags
Scribe also notes sounds that aren't speech, such as laughter or footsteps, right in the text. You'll see where people laughed, along with what they said.
Keyterm Prompting
Scribe v2 accepts a list of up to 1,000 names, brand terms or technical words, which helps it spell them right. If your recordings are full of jargon or unusual names, this is one of the ElevenLabs speech-to-text capabilities to know about.
Entity Detection
Entity detection picks out 65 types of items in the speech-to-text result, such as names and places. You can spot every person and location in a long interview at a glance.
Language Detection and Transcript Editing
You don't have to say which language is spoken, because the model detects it on its own. There's also a no-verbatim mode, plus an experimental option that edits the transcript when you describe the change in plain words.
Multichannel Audio
Did you record each mic on its own track? Scribe v2 handles files with up to 5 channels and treats each channel as a separate speaker, which keeps two-host podcasts and call recordings tidy.
How to Transcribe a File With Scribe
Open the Model Picker
Click the model picker at the bottom left of the message box. In the popup, switch to the AUD tab on the left and choose ElevenLabs Scribe from the list of audio models.
Pick a Version and Speaker Setting
On the right, pick version 1 or Latest. Latest is Scribe v2. To change the settings, click the sliders icon next to the model button, choose the version and turn Diarize on or off, then click Save. Diarize splits the text by speaker and is on by default.
Attach Your File
Click the paperclip at the bottom right of the message box and add your audio or video file. Each upload can be up to 50 MB.
Send and Keep Going
Press send, and the transcript arrives in the chat. From there, switch to any text model and ask it questions about the transcript, with the whole conversation still in view.
Scribe Languages and Accuracy
Scribe transcribes 90+ languages, according to the official list of ElevenLabs Scribe supported languages. Accuracy isn't the same across all of them. ElevenLabs puts 37 languages in its top tier, with a word error rate of 5% or lower, and places 19 more in a moderate tier. If you work in a less common language, test a short clip first.
At launch, the ElevenLabs Scribe STT model beat Whisper Large V3, Gemini 2.0 Flash and Deepgram Nova-3 on the FLEURS and Common Voice tests. ElevenLabs published this result for the first version only, so treat it as a vendor claim.
Keep Working With Your Transcript
Switch Models Without Losing the Text
Shared context keeps your conversation intact when you move from one model to another. Your transcript stays in the chat, so you can bring in Claude or ChatGPT to work with it without pasting anything again.
Add a Voice in the Same Chat
In Multi-Format mode, text and audio share one chat. Once the transcript is done, an audio model can voice your edited script. ElevenLabs voice tools and the ElevenLabs speech-to-text product are both in the same model list, so you go from recording to text to new audio without leaving the chat.
One Subscription for Every Model
You don't need a separate plan for each tool. One subscription covers every model available, including the other audio and transcription models. Scribe runs on the same tokens as everything else, so ElevenLabs transcription pricing doesn't come as a separate bill.
Use Scribe Through the API
Are you building your own tool? Scribe v1 and Scribe v2 are both in the Cabina.AI API. The MCP integration brings the same models into apps like Claude, ChatGPT and Cursor, and you can set it up with a single click or one terminal command.
Scribe Pricing on Cabina.AI
ElevenLabs Scribe pricing here depends on your plan. Each transcription costs the same number of tokens on every plan, and a bigger plan gives you more tokens each month, so more transcriptions fit in. Scribe v1 and Scribe v2 cost the same on every plan.
| Plan | Price | Tokens per month | Tokens per transcription | Transcriptions per month |
|---|---|---|---|---|
| Free | $0.00 / month | 50 | 0.36 | 138 |
| AI | $4.99 / month | 450 | 0.36 | 1250 |
| Agent | $19.99 / month | 1,000 | 0.36 | 2777 |
| Agent Pro | $49.99 / month | 2,100 | 0.36 | 5833 |
For comparison, ElevenLabs speech-to-text pricing on the Scribe API is $0.22 per audio hour for Scribe v2. That's the rate ElevenLabs charges developers directly, and it's separate from Cabina.AI plans.
ElevenLabs sells its own plans in credits and shows how much transcription time each plan includes. If you're comparing ElevenLabs speech-to-text credits per minute, Starter covers 1 hour 31 minutes a month, Creator covers 6 hours 7 minutes, and Pro covers 30 hours 18 minutes.
The ElevenLabs Scribe free limit on the ElevenLabs site is about 12 minutes of transcription a month. If you'd rather test Scribe here, the Cabina.AI Free plan lets you run it on your own recordings before you choose a paid plan.
Scribe Questions, Answered
Which ElevenLabs STT Model Should I Pick, Scribe v1 or Scribe v2?
Go with Scribe v2 for new work, since it's the current ElevenLabs STT model. Scribe v1 launched in February 2025 as the first version, and ElevenLabs now marks it as deprecated in favor of Scribe v2.
How Much Does Scribe Cost per Minute?
On Cabina.AI, there's no ElevenLabs Scribe pricing per minute. You pay a fixed number of tokens for each transcription instead.
Can I Transcribe Live Audio With Scribe?
No, the Scribe versions here transcribe recorded files. For realtime transcription, ElevenLabs offers a separate model called Scribe v2 Realtime, which returns partial results in about 150 milliseconds. That model isn't part of Cabina.AI.
What Files Can I Use, and Can Scribe Transcribe Video?
Yes, Scribe transcribes video files as well as audio. It accepts audio formats such as MP3, WAV, FLAC, M4A, AAC, OGG and Opus. For video, it takes MP4, MOV, MKV, AVI, WebM and several other common formats, so you don't need to extract the sound track first.
Is Scribe Safe for Medical or Sensitive Recordings?
Companies that need HIPAA compliance have to sign a Business Associate Agreement with the ElevenLabs sales team before they deploy Scribe. For clinical audio, ElevenLabs also makes Scribe v2 Medical, which has 35% fewer transcription errors on medical recordings than Scribe v2. That model isn't on Cabina.AI, so check your own rules before you upload patient audio.
What Are Good Alternatives to Scribe?
You can also use Whisper, GPT-4o Transcribe, GPT-4o Mini Transcribe and Cabina Transcriber for speech-to-text. They're all part of the same subscription, so you can run one file through two models and keep the transcript you like best.
Transcribe Your First File on Cabina.AI
Bring a recording and get it back as text, right next to every other model you use, all under one subscription. The Free plan lets you start without paying a cent.
Start Transcribing Free on Cabina.AI