ElevenLabs alternatives, ranked by cost per character

The table, in the one unit everybody shares

Prices are list, in USD, checked 12 September 2026. Where a figure is derived it is a plan price divided by its own published allowance, and the arithmetic is CartSignal's; the inputs are the vendors' own numbers.

Finding 1: the cheapest ElevenLabs alternative is ElevenLabs

ElevenLabs sells the same speech through two meters at very different prices. A Starter subscription costs $6 a month for 30,000 credits, and since its pricing page states text to speech burns 1 credit per character, that is an effective $200.00 per million characters. Its developer API page sells Flash and Turbo at $0.05 per 1,000 characters — $50.00 per million, exactly four times cheaper — and v3 or Multilingual v2 at $0.10 per 1,000, which at $100.00 per million still undercuts every subscription tier ElevenLabs sells, including Business billed annually at $137.50.

So a buyer whose problem is the ElevenLabs bill, and who can call an API rather than use the web app, does not need an alternative vendor. This is the second time this pattern has turned up here: the HeyGen alternatives page found that the cheapest avatar minute in the video field was HeyGen's own. In both cases the expensive thing is not the vendor, it is the packaging.

One caveat, flagged rather than buried. Multiple third-party trackers report that Flash and Turbo also burn credits at half rate — 0.5 credits per character — on the subscription meter, which would halve every ElevenLabs subscription figure above. ElevenLabs' own pricing page states 1 credit per character without naming a model, and the help-centre article that would settle it returned HTTP 403 to our fetch. The exact 2:1 ratio between v3 and Flash on ElevenLabs' official API page is consistent with the claim, so treat the subscription rows above as the worst case and check your own usage screen.

Finding 2: the volume discount stops at Pro

Climbing ElevenLabs' ladder buys less than it looks. Per million characters, monthly billing runs $200.00 on Starter, $181.82 on Creator, then $165.00 on Pro — and then it simply stops. Business at $990 a month buys ten times Pro's credits at exactly the same $165.00 per million. Scale, the tier in between, is the only one that moves in the wrong direction: at $166.11 per million it costs about 0.7% more per character than the tier below it.

The practical reading is that ElevenLabs' unit price bottoms out at Pro, so the reason to buy Scale or Business is seats, support and contractual terms, not cheaper speech. Annual billing is the discount that actually compounds — $137.50 per million on Pro and Business — but that is still 2.75x ElevenLabs' own Flash API rate.

Finding 3: the field bills in four incompatible units

This is the trap underneath every "cheapest TTS API" listicle. Four different units are in play, and only the first is directly comparable:

• Per character. ElevenLabs, Deepgram, Murf's API, Speechify, Amazon Polly and OpenAI's tts-1 and tts-1-hd. These are the rows in the table above.
• Per credit, with a character equivalence. Cartesia, whose plans imply a flat 750 credits per minute — its free tier's 20,000 credits buy a stated ~27 minutes (740/min) and Scale's 8M buy ~10,667 (exactly 750/min).
• Per audio output token. OpenAI's gpt-4o-mini-tts at $12.00 per million audio output tokens, gpt-realtime-mini at $2.40 and gpt-audio at $64.00; Google's Gemini 2.5 Flash Preview TTS at $10.00 per million output tokens and Gemini 2.5 Pro Preview TTS at $20.00. No official page converts audio tokens to characters, so these cannot be placed in the table, and a $10.00-per-million token headline is not cheaper than a $15.00-per-million character one — it is a different quantity.
• Per minute. Murf's Falcon 2 at $0.01 per minute, Cartesia's voice agents at $0.06, and ElevenLabs' non-TTS products — music $0.15, dubbing $0.33 with watermark rising to $2.20 on v2, voice changer $0.12.

Ranking by dollars per million characters is the fix, because it does not depend on how fast a voice speaks. Ranking by dollars per minute does, and that is where comparison pages go wrong — including, until today, this one.

A correction to this page's own arithmetic

Earlier versions of this page and of the voice comparison said Cartesia cost about $0.038 per minute against roughly $0.18 on ElevenLabs Starter, "roughly 5x". That gap was inflated by a unit slip: 750 characters per minute was applied to Cartesia and about 900 to ElevenLabs. On any single density constant the ratio is exactly 4.0x ($50.00 against $200.00 per million), and at one consistent 750 characters per minute the two work out at about $0.0375 and $0.150 a minute. Cartesia is still materially cheaper; it is not five times cheaper.

Finding 4: what the 50x actually buys

A 50x price gap between Amazon Polly Standard and ElevenLabs Starter is real, but it is not a 50x saving on the same thing, and it would be dishonest to present it as one. Polly's Standard voices are its oldest and least natural tier. ElevenLabs' price includes voice cloning, emotional direction, dubbing, sound effects and music from a single pool, and it is the reference point the rest of the field is benchmarked against for realism.

The comparison that respects that is like-for-like across the premium end: ElevenLabs v3 API at $100.00 per million, Polly Long-form at $100.00, Polly Generative and Deepgram Aura-2 and Murf at $30.00, OpenAI tts-1-hd at $30.00, Cartesia at $37–50. That is a 3.3x band, not 50x — the same shape the video field showed, where premium output has largely commoditised and the extreme ends of the range are different products rather than different prices.

Amazon Polly is worth a line of its own here, because a single vendor spans 25x internally: $4.00 Standard, $16.00 Neural, $30.00 Generative, $100.00 Long-form. Note the ordering — Long-form costs 3.3x the newer Generative tier. As on Fliki and Seedance, the newest model is not reliably the dearest one, so read the rate card rather than assuming.

Finding 5: Speechify's ladder runs backwards

Speechify's API pricing page lists Starter at $10 a month for 1.9M characters, Pro at $99 for 13.5M, and Scale at $499 for 78M. Divided out, that is $5.26, $7.33 and $6.40 per million — the entry tier is the cheapest per character and Pro is the most expensive, a 39% penalty for moving up one rung. Every bundled rate also undercuts that tier's own overage rate ($10, $8 and $6 per million), so on this ladder the included characters are always the good ones.

Two cautions. Third-party trackers report a different allowance ladder for the same tiers (1M, 3M and 10M characters), which would change these figures substantially; the numbers above are what the official page returned on 12 September 2026, and the Speechify profile carries the fuller picture of its three separate meters. And Speechify's page states plainly that voice agents are not on this ladder and carry no published per-minute rate at all.

Who does not publish a rate you can read

Two of the three hyperscalers failed the simplest test — printing a number. Azure's Speech pricing page renders every text-to-speech cell as "$-" until a region is selected, so no rate was machine-readable. Google Cloud Text-to-Speech truncated its pricing table on fetch. Both are credible ElevenLabs alternatives and both are listed here without a figure rather than with a guessed one — the same treatment Fliki and VEED get on this site for rendering their prices client-side.

Two names to strike off the shortlist

Older ElevenLabs-alternative roundups still recommend two products that cannot be bought:

• PlayHT. Meta acquired the team in July 2025, the TTS service was deprecated within weeks, and the domain no longer resolves. The migration page has the sourced timeline.
• Resemble AI. Resemble said on 13 August 2026 it is "not selling voice AI to new customers" and moved to deepfake detection; its pricing page carries no TTS rate. Its models do remain free and self-hostable as MIT-licensed Chatterbox, which is the genuine zero-cost option in this field.

How to choose

• Price the meter, not the vendor. ElevenLabs' API is 4x cheaper than ElevenLabs' entry subscription for identical speech. Check whether your own vendor has a second, cheaper meter before you migrate.
• Convert to characters before comparing. Minutes depend on speaking density and tokens are not characters. A headline rate in a different unit is not a cheaper rate.
• Buy the tier, then check it is not the bad rung. ElevenLabs Scale costs more per character than Pro; Speechify Pro costs more than Starter.
• Match the quality tier, not the floor. Polly Standard at $4.00 is not a substitute for ElevenLabs v3. Compare premium against premium and the field narrows to about 3.3x.
• Check what you lose besides realism. Cloning, dubbing, music and an editor are bundled into ElevenLabs' price. Murf is the workflow-first answer, Cartesia the low-latency one.
• If you publish into the EU, synthetic speech falls under Article 50 disclosure duties whichever engine made it, and a vendor watermark does not discharge your own duty.

Related reading

Individual rate-card breakdowns are in the ElevenLabs, Cartesia, Speechify and Murf profiles. The head-to-head is ElevenLabs vs PlayHT, the curated shortlist is best AI voice tools, and the wider directory is the AI voice category. For the same analysis applied to avatar video, see HeyGen alternatives.

FAQ

What is the cheapest alternative to ElevenLabs?

On published list rates in September 2026, Amazon Polly's Standard voices at $4.00 per million characters are the cheapest way to turn text into speech, about 50 times cheaper than an ElevenLabs Starter subscription at an effective $200.00 per million. But the closest like-for-like saving needs no vendor switch at all: ElevenLabs' own API prices Flash and Turbo at $0.05 per 1,000 characters, which is $50.00 per million — four times cheaper than its own Starter plan, and cheaper than every subscription tier it sells.

How much does ElevenLabs actually cost per character?

ElevenLabs' official pricing page states text to speech costs 1 credit per character, so dividing each plan price by its credit allowance gives the effective rate directly. Starter is $200.00 per million characters, Creator $181.82 monthly or $151.49 billed annually, Pro $165.00 monthly or $137.50 annually, Scale $166.11 monthly, and Business $165.00 monthly or $137.50 annually. The discount stops at Pro: Business buys ten times the credits at exactly the same unit price.

Why do text to speech prices per million characters not compare cleanly?

Because the field bills in four incompatible units. ElevenLabs, Deepgram, Murf, Speechify, Amazon Polly and OpenAI's tts-1 models bill per character. Cartesia bills credits with a fixed 750-credits-per-minute equivalence. OpenAI's gpt-4o-mini-tts and Google's Gemini TTS models bill per million audio output tokens, which no official page converts to characters. Murf's Falcon 2 and Cartesia's voice agents bill per minute. A per-token headline of $10 per million cannot be compared with a per-character headline of $15 per million.

Is Cartesia five times cheaper than ElevenLabs?

No, it is four times cheaper, and the difference is a measurement error this site previously published. Cartesia Pro works out at $50.00 per million characters against ElevenLabs Starter at $200.00, which is exactly 4.0x. An earlier version of this page reported roughly 5x because it applied 750 characters per minute to Cartesia and 900 to ElevenLabs. Ranking by dollars per million characters avoids the problem entirely, because it does not depend on an assumed speaking density.

Is this a hands-on test of voice quality?

No. This is a public-data price comparison built from official pricing pages and rate cards checked on 12 September 2026. It compares what each vendor charges per character of synthesised speech, not how the speech sounds. A voice that costs 50 times less is not the same product, and no listening, prosody or cloning-quality testing is claimed here.

Source links

• Official ElevenLabs pricing page (plan prices, credit allowances, 1 credit per character) — checked 12 September 2026
• Official ElevenLabs API pricing (per-model rates per 1,000 characters, dubbing, music, voice changer)
• Official Cartesia pricing page (plan prices, credit allowances, included minutes, agent rates)
• Official Speechify API pricing page (tier prices, bundled characters, overage rates)
• Official Deepgram pricing page (Aura-1 and Aura-2 per-1,000-character rates)
• Official OpenAI API pricing (tts-1 and tts-1-hd per million characters, audio-token model rates)
• Official Amazon Polly pricing (Standard, Neural, Generative and Long-form per million characters)
• Official Gemini API pricing (TTS model output-token rates)
• CartSignal Murf profile — carries the Murf API and Falcon 2 rates and links to the official pages they come from