# ElevenLabs Scribe Alternative for Voice Agents | AirCaps Live 1

URL: https://aircaps.ai/compare/elevenlabs-scribe-alternative

> AirCaps Live 1 vs ElevenLabs Scribe v3 Live: 38 vs 79 ms to final transcript, 3.65% vs 4.10% WER, $0.20 vs $0.22 per hour. Built for voice agents.

[AirCaps Live 1](https://aircaps.ai/models/live-1) ElevenLabs alternative

# An ElevenLabs Scribe alternative *for voice agents.*

AirCaps Live 1 returns the final transcript a median 38 ms after the caller stops speaking, against 79 ms for ElevenLabs Scribe v3 Live, and makes 11% fewer word errors (3.65% against 4.10% WER on 9,205 voice-agent clips). It also detects the end of each turn itself, which ElevenLabs Scribe v3 Live does not. It costs $0.20 per audio hour; ElevenLabs Scribe v3 Live lists at $0.22.

[Start free](https://console.aircaps.ai/) [See AirCaps Live 1](https://aircaps.ai/models/live-1)

**38 ms** median time to final transcript; Scribe v3 Live takes 79 ms, and 2.3× as long at p95

**3.65%** word error rate on 9,205 voice-agent clips; Scribe v3 Live 4.10%

**Built in** turn detection, which Scribe v3 Live does not have

**$0.20** per audio hour; Scribe v3 Live lists at $0.22

## Side by side.

The same audio, streamed in real time to every model. The better result in each row is highlighted.

| Lower is better, except where noted | AirCaps Live 1 | ElevenLabs Scribe v3 Live | ElevenLabs Scribe v2 Realtime |
|---|---|---|---|
| Word error rate, all 9,205 clips | 3.65% | 4.10% | 4.48% |
| Coval public STT set | 1.12% | 1.32% | 1.41% |
| Pipecat STT benchmark | 1.52% | 1.70% | 1.85% |
| LiveKit EOT Bench | 7.02% | 9.02% | 8.86% |
| LiveKit EOT Bench, accented English | 5.23% | 5.85% | 6.11% |
| Time to final transcript, median | 38 ms | 79 ms | 101 ms |
| Time to final transcript, p95 | 52 ms | 121 ms | 146 ms |
| Built-in turn detection | Yes | No | No |
| List price per audio hour | $0.20 | $0.22 | $0.22 |

9,205 English clips (154,661 words) from Pipecat's STT benchmark, Coval's public set, LiveKit EOT Bench, WildASR and PriMock57, run on November 19, 2026, from one machine in AWS us-east-2. Time to final transcript runs from the end of speech, when the last audio and the finalize message have been sent, to the final transcript, network included. Turn detection: Pipecat's Smart Turn v3.2 English set (5,367 clips) and LiveKit EOT Bench English (7,655 turns), each model's own turn detection. Prices are list prices per hour of English streaming audio.

## What changes when you switch.

### Fewer word errors

3.65% WER on 9,205 voice-agent clips, against 4.10% and 4.48% for Scribe v3 Live and Scribe v2 Realtime. Live 1 is also lower on each of the four public sets in the table.

### A sooner final transcript

A median 38 ms from the end of speech to the final transcript, 52 ms at p95. Scribe v3 Live takes 79 ms, and 121 ms at p95.

### Turn detection built in

Live 1 decides when the caller has finished, from what they said and how long they paused. With Scribe v3 Live and Scribe v2 Realtime you run a separate VAD or turn model; with Live 1 you can still send your own `finalize`.

### A lower price

$0.20 per audio hour, with 5 free hours on every account. Scribe v3 Live lists at $0.22.

## Questions.

**Is AirCaps Live 1 a good alternative to ElevenLabs Scribe v3 Live?**

AirCaps Live 1 returns the final transcript a median 38 ms after the caller stops speaking, against 79 ms for ElevenLabs Scribe v3 Live, and makes 11% fewer word errors (3.65% against 4.10% WER on 9,205 voice-agent clips). It also detects the end of each turn itself, which ElevenLabs Scribe v3 Live does not. It costs $0.20 per audio hour; ElevenLabs Scribe v3 Live lists at $0.22.

**How do I switch from ElevenLabs to AirCaps Live 1?**

Stream 16 kHz mono PCM to the Live 1 WebSocket and read the transcript messages. Turn on end_of_turn to let the model end turns, or send finalize from your own VAD; both work in the same session. The AirCaps Live 1 guide at aircaps.ai/docs/guides/live-1 has the whole protocol.

**How was this measured?**

9,205 English clips (154,661 words) from Pipecat's STT benchmark, Coval's public set, LiveKit EOT Bench, WildASR and PriMock57, run on November 19, 2026. Every model streamed the same audio in real time from one machine in AWS us-east-2. Time to final transcript runs from the end of speech, when the last audio and the finalize message have been sent, to the final transcript, network included.

**How much does AirCaps Live 1 cost?**

$0.20 per hour of audio you stream, billed per millisecond, with 5 free hours on every account. ElevenLabs Scribe v3 Live lists at $0.22; ElevenLabs Scribe v2 Realtime lists at $0.22.

**Does it work with Pipecat and LiveKit?**

Yes. It runs over a WebSocket, so it fits a Pipecat pipeline or a LiveKit agent as the speech-to-text step.

**Which languages does it support?**

English, with multilingual support coming soon.

## More comparisons.

[**Deepgram alternative** Live 1 vs Flux and Nova-3 *Flux*](https://aircaps.ai/compare/deepgram-flux-alternative) [**AssemblyAI alternative** Live 1 vs Universal-3.6 Pro and Universal-3.5 Pro *Universal-3.6 Pro*](https://aircaps.ai/compare/assemblyai-alternative) [**AirCaps Live 1** All benchmarks *$0.20/h*](https://aircaps.ai/models/live-1)

## Try it on *your calls.*

AirCaps Live 1 is on every account, with 5 free hours. Read the [AirCaps Live 1 guide](https://aircaps.ai/docs/guides/live-1).

[Start free](https://console.aircaps.ai/) [Talk to sales](https://aircaps.ai/contact)
