ElevenLabs API pricing: TTS, sound effects, and speech to text

ElevenLabs is the audio suite most developers reach first: text-to-speech, sound effects, transcription, multi-voice dialogue, voice conversion, and vocal isolation. On UnificAlly, all six tools run through one POST /v1/tasks endpoint with one API key, priced per generation with no subscription. This post covers what each tool does, what it costs, and how to call it.

TL;DR: Six ElevenLabs tools on one API: TTS from $0.024 per 1,000 characters (4 model variants, 18 output formats), Sound Effects at $0.0194 per effect, Speech-to-Text with diarization and word-level timestamps, Dialogue at $0.048 per 1,000 characters, Voice Changer at $0.0968 per minute, and Voice Isolation. Pay per generation, no monthly fee, playground on every page.

What the ElevenLabs API includes

Every tool returns a task_id you poll, or a webhook callback if you pass callback_url. Same request shape across all six.

ElevenLabs API pricing

You pay per finished generation. No subscription, no monthly minimum, no credit packs that expire.

Three practical notes on cost:

• Sound effects are flat-rate per effect, whether the result is half a second or 30 seconds.
• Narrating a 500-word article (about 3,000 characters) with Flash costs $0.072.
• There is no free tier, but the ElevenLabs API playground lets you test every tool from the browser before you write any code.

ElevenLabs TTS API: models, voices, and formats

The TTS endpoint exposes four ElevenLabs models. Pick by latency, language coverage, and how much text you need per call:

eleven_flash_v2_5 is the default. Voices come from GET /v1/resources/elevenlabs/voices, which lists the full ElevenLabs voice library, including any cloned voices on your account, so custom voices show up without redeploying anything.

Output formats cover 18 options: MP3 at five bitrates, Opus at five, raw PCM at five sample rates, WAV 44.1 kHz, and the telephony codecs ulaw_8000 and alaw_8000 for Twilio-style integrations. A voice_settings object controls stability, similarity, style, and speed, and a seed makes output reproducible.

ElevenLabs Sound Effects API

The sound effects endpoint turns a text description into an audio asset: "wolf howling at the moon", "rain falling on a window", "busy coffee shop with people chatting". At $0.0194 per effect, it is cheaper than licensing a stock SFX pack and faster than recording foley.

Controls that matter:

• duration: auto (default) or a fixed 0.5–30 seconds.
• loop: make the audio loop cleanly, for ambient beds and game audio.
• prompt_influence (0–1): how literally the model follows your description.

Game one-shots, UI feedback sounds, podcast stingers, and ambient loops are the sweet spot. For music with vocals, use Suno instead; our Suno pricing guide covers what songs cost per request.

ElevenLabs Speech-to-Text API (transcription)

The STT endpoint runs ElevenLabs Scribe v2. Point it at an audio_url and it returns the transcript plus a word array with start and end times. It goes well past plain transcription:

• Speaker diarization: who said what, up to 32 speakers.
• Timestamps: per word or per character, for captions and subtitles.
• Audio event tags: (laughter), (music), (footsteps) inline in the transcript.
• Entity detection: flag PII, PHI, PCI, or offensive language.
• Key terms: up to 100 names or jargon words to boost recognition accuracy.
• Language: auto-detect, or force one of the supported ISO codes.

Meeting notes, podcast transcripts, compliance review, and subtitle files all come out of the same call.

Dialogue, voice changer, and voice isolation

Text-to-Dialogue

The dialogue endpoint generates a full multi-speaker conversation in one call. Each line carries its own voice ID and optional emotion tag: [excitedly], [whispered], [sadly]. It always runs on eleven_v3, and bills $0.048 per 1,000 characters. Podcast intros, game NPC banter, and audio drama scenes are the obvious fits.

Voice Changer

The voice changer converts the voice in an existing recording to any target ElevenLabs voice while keeping the timing and inflection of the original performance. Optional background noise removal cleans the input first. $0.0968 per minute of processed audio.

Voice Isolation

The isolation endpoint takes one parameter, audio_url, and returns the vocal with music and noise stripped out. Use it to rescue location recordings or print a dry vocal stem for editing.

How to call the ElevenLabs API

Every tool uses the same request shape. Here is TTS:

curl -X POST https://api.unifically.com/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "elevenlabs/text-to-speech",
    "input": {
      "text": "Hello world, this is a test.",
      "voice": "TX3LPaxmHKxFdv7VOQHJ",
      "model_id": "eleven_flash_v2_5",
      "output_format": "mp3_44100_128"
    }
  }'

The response returns a task_id. Poll GET /v1/tasks/<task_id> until status is completed and read output.audio_url, or pass a callback_url and receive the result as a webhook. Swap the model field to elevenlabs/sound-effect, elevenlabs/speech-to-text, or any of the other IDs and change the input accordingly. Full parameter docs live at docs.unifically.com.

Which ElevenLabs model should you use?

• Voice agents and live apps: eleven_flash_v2_5. At ~75ms it is the only one fast enough for real-time back-and-forth.
• Audiobooks and narration: eleven_multilingual_v2 for the richest delivery, or eleven_v3 when you want dramatic range and 70+ languages.
• Bulk content at the best price-to-quality: eleven_turbo_v2_5, with the same 40,000-character limit as Flash.
• Scripted scenes with several characters: skip stitching TTS calls and use Dialogue directly.
• Cleaning up real recordings: Voice Isolation first, then Voice Changer if the voice itself needs to change.

ElevenLabs FAQ

Is this the official ElevenLabs text to speech pricing for 2026?

ElevenLabs (often written Eleven Labs) publishes its official subscription plans on its own site. The rates on this page are UnificAlly's per-generation API rates, refreshed from our live price list on every deploy. If you want the ElevenLabs API price without a subscription, this table is the current answer.

Is there a free ElevenLabs API key?

No. On UnificAlly you pay per generation with no monthly fee, and the ElevenLabs playground is the quickest way to test every tool before integrating.

How much does the ElevenLabs API cost?

TTS is $0.024 per 1,000 characters on Flash and Turbo ($0.048 on Multilingual v2), sound effects are $0.0194 per effect, dialogue is $0.048 per 1,000 characters, and voice changer is $0.0968 per minute. Speech-to-text and isolation rates are on the pricing page. Everything bills per finished generation.

Can I use my cloned ElevenLabs voices?

Yes. GET /v1/resources/elevenlabs/voices returns the voice library including cloned voices tied to the platform account, and any voice ID from that list works in TTS, dialogue, and voice changer calls.

Does the ElevenLabs API do transcription?

Yes. elevenlabs/speech-to-text transcribes audio with speaker diarization (up to 32 speakers), word-level timestamps, audio event tags, and PII/PHI/PCI entity detection.

What audio formats does TTS output?

18 formats: MP3 (five bitrates), Opus (five), PCM (five sample rates), WAV 44.1 kHz, and telephony ulaw_8000 / alaw_8000.

All six tools are live on the ElevenLabs model page with a playground for each. We will update this post when ElevenLabs ships new TTS model versions or when per-tool rates change.