ElevenLabs Forced Alignment: How It Works & Costs (2026)

ElevenLabs Forced Alignment is an API-only tool that aligns your text with spoken audio to produce word-level timestamps. It supports 29 languages, accepts files up to 3 GB or 10 hours, and bills at the same rate as the Speech to Text API. Checked 5 Nov 2026.

What is ElevenLabs Forced Alignment?

ElevenLabs Forced Alignment takes two inputs — an audio file containing speech and the exact text that was spoken — and returns a transcript where every word carries its start and end time. Unlike speech-to-text, which guesses the words from audio alone, forced alignment starts from text you already trust and focuses purely on timing. The result is a word-level map of your audio: you know precisely when each word begins and ends.

The feature sits in the ElevenLabs API under the Speech to Text capability family, but it solves a different problem. Speech-to-text answers “what was said?” Forced alignment answers “when was each word said?” That distinction matters for anything where timing drives the experience: subtitles that must land on the right frame, audiobook players that highlight the sentence being read, or language apps that let learners tap a word to hear it in context.

ElevenLabs documents two headline use cases: matching subtitles to a video recording, and generating timings for an audiobook recording of an ebook. Both follow the same pattern — you hold the script, you hold the recording, and you need them locked together in time. The API returns the aligned words with timestamps you can feed straight into subtitle files, ebook narration tracks, or interactive players.

One important boundary: Forced Alignment is API-only. There is no button for it in the ElevenLabs dashboard. You interact with it through the ElevenLabs API directly, which means you need an API key and the ability to make HTTP requests. If you need a key, our guide to getting an ElevenLabs API key walks through the exact steps.

How to use Forced Alignment in 5 steps

Step 1: Get an ElevenLabs API key. Create an ElevenLabs account, open your profile settings, and generate an API key. The Free plan includes API access with 10,000 credits per month, which is enough to test alignment on short files before committing to a paid tier. Cost: 0 credits to generate the key.

Step 2: Prepare your audio file. Forced Alignment accepts audio up to 3 GB in size and up to 10 hours in duration. Supported formats include MP3, WAV, FLAC, M4A, OGG, OPUS, AIFF, AAC, and WebM; video containers like MP4 and MOV are also accepted because the API extracts the audio track. For best results, use clear recordings with minimal background noise, since alignment accuracy depends on the audio matching the text closely. Cost: 0 credits to prepare.

Step 3: Prepare your text as a plain string. The input text must be a plain string — no SSML, no audio tags, no HTML markup. The maximum text length is 675,000 characters, which covers very long documents of 100,000 words or more. Make sure the text matches what was actually spoken, word for word. If the speaker ad-libbed, skipped sentences, or the transcript contains errors, alignment quality drops because the engine tries to map every word you provided. Cost: 0 credits to prepare.

Step 4: Call the Forced Alignment endpoint. Send a POST request to the ElevenLabs API with your audio file and text string, authenticating with your API key in the xi-api-key header. The API processes the file and returns a JSON response containing each word with its start and end times. Larger files take longer to process, so build your integration to handle asynchronous responses for long audio. Cost: billed at the same rate as the Speech to Text API (checked 5 Nov 2026) — you pay for the audio processed, not for the text length.

Step 5: Use the timestamps in your product. The response gives you word-level timing you can convert into SRT or VTT subtitle files, chapter markers, audiobook narration tracks, or karaoke-style highlighting. Because the output is structured JSON, you can transform it programmatically into whatever format your player needs. Cost: 0 credits to transform the output — you only paid for the alignment call in Step 4.

Limits and technical requirements

Forced Alignment has hard limits you should design around. The maximum file size is 3 GB and the maximum audio duration is 10 hours, which comfortably covers full-length audiobooks and feature films. The text cap of 675,000 characters means you can align very long scripts in a single call instead of chunking them manually.

Language support covers 29 languages through the multilingual v2 models. If your audio is in one of those 29 languages and your text matches the recording, alignment works. If your language sits outside that set, the API cannot align it, so check the supported-language list in the docs before building.

Two constraints catch people out. First, the input text must be a plain string. You cannot pass SSML, audio tags like [laughing], or HTML — strip all markup before sending. Second, speaker diarization is not supported. If your audio has multiple speakers, the API returns word timings but does not label who spoke which word.

Forced Alignment vs Speech to Text

These two APIs sound similar but answer different questions. Speech to Text listens to audio and guesses the words — use it when you do not have a transcript. Forced Alignment takes audio plus a transcript you trust and computes when each word occurs — use it when you hold the script and need timing.

If you are building subtitles for a video where you already have the script, Forced Alignment is the direct tool — it is faster and more accurate for timing because it never guesses words. If you recorded a meeting and need to know what was said, Speech to Text with the Scribe v2 model is the right call. Some pipelines use both: Scribe v2 generates a draft transcript, a human corrects it, then Forced Alignment locks the corrected text to the audio with precise timing.

Real-world use cases

The docs name two primary uses, and both are substantial. First, subtitle timing for video: with a recording and its script, Forced Alignment produces the word timings needed for perfectly synced subtitles or captions, including word-by-word karaoke highlighting. Second, audiobook production: publishers holding an ebook text and a narrated recording can generate the timing map that powers read-along highlighting in audiobook apps.

Beyond those, developers use word timestamps for language-learning apps where tapping a word plays it back, podcast chapter markers derived from a script, video editing tools that cut to the word, and accessibility features like precise caption syncing. Anywhere the text is known and the timing is the missing piece, forced alignment fits.

Who should NOT use it

Skip Forced Alignment if you do not have a transcript — use Speech to Text instead. Skip it if you need speaker labels on multi-speaker audio, since diarization is not supported. Skip it if you want a no-code tool, because there is no dashboard interface. And skip it if your language falls outside the 29 supported languages. For everyone else holding a script and a recording, it is the most direct path to word-level timing.

Frequently asked questions

What is ElevenLabs Forced Alignment?

It is an API that aligns a known text with spoken audio to produce word-level start and end timestamps. You provide the audio file and the exact transcript; the API returns when each word was spoken. It supports 29 languages, files up to 3 GB or 10 hours, and text up to 675,000 characters. Checked 5 Nov 2026.

How is Forced Alignment different from Speech to Text?

Speech to Text guesses the words from audio alone, while Forced Alignment starts from text you provide and computes only the timing. Use speech-to-text when you need a transcript; use forced alignment when you already have the script and need word-level timestamps for subtitles, captions, or audiobooks.

How much does ElevenLabs Forced Alignment cost?

It bills at the same rate as the Speech to Text API — you pay for the audio processed, not the text length. The Free plan includes 10,000 credits per month (checked 5 Nov 2026), and Starter is $6/mo with 30,000 credits (checked 5 Nov 2026). Check the official API pricing page for current per-minute rates before building.

What are the file limits for Forced Alignment?

Maximum file size is 3 GB, maximum audio duration is 10 hours, and maximum text length is 675,000 characters. Supported audio formats include MP3, WAV, FLAC, M4A, and OGG; video files like MP4 are accepted with the audio extracted. The input text must be a plain string with no markup.

Does Forced Alignment support speaker diarization?

No. Forced Alignment returns word-level timestamps but does not label which speaker said each word. If you need speaker labels on multi-speaker audio, use the Speech to Text API with the Scribe v2 model and diarization enabled, which supports up to 32 speakers.

Can I use Forced Alignment without coding?

No. Forced Alignment is API-only with no dashboard interface. You need an API key and the ability to make HTTP requests. Non-developers should use the Speech to Text dashboard tools or a no-code platform that wraps the ElevenLabs API.

Sources checked

• ElevenLabs documentation: Forced Alignment — https://elevenlabs.io/docs/overview/capabilities/forced-alignment (checked 5 Nov 2026)
• ElevenLabs documentation: Models — https://elevenlabs.io/docs/overview/models (checked 5 Nov 2026)
• ElevenLabs pricing — https://elevenlabs.io/pricing (checked 5 Nov 2026)