Speechify public-data review: three products, three meters

Three products, and which one you actually want

Almost every confusing thing written about Speechify pricing comes from treating it as one product. It is three, on three pricing pages, with three billing models and no shared allowance between them.

The trap is the first row. Search for "Speechify pricing" and you land on the reading-app page, where Premium is $29 a month — $348 a year. Studio Starter, the tier that actually generates commercial voiceovers, is listed at $100 a year. The reading app costs roughly 3.5x the product that does the work, and buys none of it.

Speechify Studio: the plans and what a credit buys

Speechify's official Studio pricing page showed annual pricing only when checked on 10 September 2026.

And the rate card, published on the same page:

The 30x gap that decides which plan you need

One pool, three rates, and a 30x spread between the cheapest and dearest. That means a plan's headline credit number has no single meaning in minutes — it depends entirely on what you generate.

Read the Starter row across: the same $100 is either 24 hours of narration or 48 minutes of avatar video for the entire year. Anyone buying Studio for avatar work should size the plan against 1,800 credits a minute and ignore the voiceover framing completely.

Cost per finished minute

Creator is exactly 25% cheaper per credit than Starter — 3x the price for 4x the credits, which divides out to precisely 0.75. That is worth noting because it runs the opposite way to Synthesia, where the Creator tier costs about 17% more per credit than Starter on yearly billing, and to HeyGen, where Business costs roughly twice per credit what Pro does. On this site's shelf, Speechify Studio is one of the few ladders where climbing it actually gets cheaper per unit.

The avatar figure is also the one that puts Speechify into the video conversation. At about $1.56 per avatar minute, Studio Creator undercuts Synthesia Starter on yearly billing, which works out near $1.80 a minute at its flat 120 credits per minute. That is a list-price comparison between two published rate cards, not a quality judgement — this review has not compared avatar output.

On the annual-only price

Speechify's Studio page listed $100 and $300 per year and no monthly figure during this review. A number of third-party pricing trackers report Studio Starter at $19 a month and Studio Creator at $49, with 7,200 and 28,800 credits respectively. Those monthly credit figures are exactly one twelfth of the annual allowances Speechify publishes (7,200 x 12 = 86,400; 28,800 x 12 = 345,600), which corroborates that monthly tiers exist and are metered consistently. The prices are not quoted from a Speechify page here. If they are right, annual billing is a 56% discount on Starter and 49% on Creator — steeper than any other annual discount costed out on this site, and steep enough to be worth confirming at checkout rather than assuming. Trackers also variously describe Speechify as having eight plans; only three appeared on the official Studio page.

One genuinely generous rule, and it is the mirror image of Descript: Speechify meters generation, not export. Re-exporting a project you have already generated costs nothing, so revising a cut and re-rendering is free. Descript, by contrast, bills the footage going in.

The API is about ten times cheaper than the Studio

Speechify's developer pricing page is a different meter entirely, and a far cheaper one. Text to speech is billed by characters sent to the API, with whitespace and SSML tags not counted. Voice agents are billed per minute, all-inclusive of the language model, speech to text, text to speech and orchestration.

Speechify states these allowances as approximations, and each plan's characters and agent minutes appear to be alternative ways of spending one bundled balance rather than two separate pots. The arithmetic below uses the published figures as printed.

Starter costs exactly what the free tier costs

Divide each plan's price by its included characters and compare that to its own overage rate. Starter's bundled characters work out at $6.67 per million against a $10 overage — a third cheaper. Pro's are $7.62 against $8, barely cheaper at all. Scale's are $6.65 against a $6 overage, meaning the characters built into the subscription cost about 11% more than the ones you buy on top of it. That is an inversion worth knowing before over-buying the tier.

Starter is stranger still. It adds 1M characters over the free tier and charges $10 for them, while both tiers carry the identical $10 per million overage rate. Work the totals through and the two plans cost the same amount at every volume of 1.5M characters or more — 3M characters is $25 either way, 5M is $45 either way — and the free tier is cheaper below that. On the published numbers Starter buys nothing. This assumes the free tier genuinely permits metered overage, which the pricing page implies by printing an overage rate against it, and it ignores any rate-limit or support differences Speechify does not publish. Confirm both before planning around it.

The tiers that do pay off are the big ones. Against Starter-plus-overage, Pro wins above about 10.4M characters a month and Scale wins above about 63M.

Cost per spoken minute, against the rest of the shelf

To compare a character meter with a per-minute one you need a script density. This review uses 750 billable characters per spoken minute — roughly 150 words a minute at five characters a word, with whitespace excluded because Speechify does not count it. That is the same assumption Cartesia's own advertised minutes imply, which makes the two directly comparable.

Two things fall out. First, Speechify's own Studio costs about 10x its own API for the same minute of generated speech. That gap is the price of the interface — Studio bundles a timeline editor, stock media, cloning, dubbing and avatars, and the API returns raw audio — but it is a large premium to pay unknowingly, and it is far wider than Synthesia's API, which undercuts its own editor by 37.5%.

Second, on list price the Speechify API is the cheapest text-to-speech meter costed out on this site: roughly 7.6x below Cartesia and 36x below ElevenLabs Starter. Treat that as a price finding only. Cartesia sells documented latency and concurrency guarantees, ElevenLabs sells voice realism and cloning depth, and CartSignal has compared none of their audio. A price this far below the field is a reason to run your own listening test, not a reason to skip one.

Voice cloning consent: a complaint, and a change

This is the part of Speechify a buyer should read before the pricing. On 27 July 2026 the Consumer Federation of America urged the FTC and state attorneys general to investigate Speechify, alleging practices that violate Section 5(a) of the FTC Act. The complaint says Speechify's cloning "enables these scams by allowing users to create realistic voice clones from just seconds of audio while relying on little more than a self-attestation checkbox to verify consent", and quotes Ben Winters, its Director of AI and Data Privacy, saying the company "has made it remarkably easy for anyone to clone a person's voice while doing almost nothing to verify consent or prevent abuse". CFA framed it against the FTC's figure of $3.5 billion in impersonation-fraud losses in 2025. These are allegations in a complaint, not findings; no regulatory outcome was public at the time of this review.

Roughly two and a half weeks later, an entry in Speechify's developer changelog filed under 13 August 2026 replaced the old consent object with a challenge-response flow. You call POST /v1/voices/consent-challenges with the speaker's full name, Speechify returns a phrase and an id, you show the speaker that exact phrase and record them reading it aloud, then submit the recording with the challenge id. Speechify transcribes it, checks it against the phrase it issued, and keeps it as the consent record for that voice.

That is a real tightening: an issued-phrase recording is materially harder to fake than a checkbox. Two honest caveats. CartSignal has not tested the flow, and does not assert the complaint caused the change — the two are dated facts in sequence, nothing more. And the changelog documents the API; whether the Studio cloning UI enforces the same check was not established from a public page.

Rights, disclosure and the EU AI Act

Speechify's Studio terms are unusually explicit on three points that matter commercially.

• You own the output. The terms state Speechify claims "no Intellectual Property Rights over the Files you Upload or the Content you Generate", and that "Your Content is yours".
• Commercial use is permitted provided you hold the necessary rights in the content you generate — and, per the pricing page, provided you are on a paid tier. The free plan carries no commercial rights.
• You must disclose. Users attest they "will include or embed commercially reasonable disclosures clearly indicating that the Synthetic Output is AI-generated and not a human voice." Cloning attestations additionally require that the voice is yours or covered by explicit written consent, that the speaker is over 18, and that the voice does not belong to a current or former reasonably well-known political figure.

That disclosure clause pushes the marking duty onto you contractually, which lines up with the provider-versus-deployer split explained in our EU AI Act labeling guide: since 2 August 2026, Article 50 requires generative audio to be machine-readably marked and requires deployers to disclose synthetic audio that could pass as authentic. This review found no official Speechify statement that it machine-marks or watermarks generated audio at the provider end. That is an open question rather than a finding, and it is the specific thing an EU publisher should ask before buying — ElevenLabs, by contrast, began SynthID audio watermarking in June 2026.

Best-fit use cases

• High-volume programmatic narration, where the API's per-character rate is the cheapest listed here by a wide margin
• Creators who want one browser tool covering voiceover, dubbing and avatar video without stitching three subscriptions together
• Video localisation on a budget, at roughly $0.16 per dubbed minute on Studio Creator
• Voice agents wanting one all-inclusive per-minute price rather than separately metered model, transcription and speech costs
• Teams that iterate heavily on edits, since re-exporting already-generated content is free

Limitations to verify

• Monthly Studio pricing was not readable on an official page; only annual $100 and $300 were listed
• Credit rollover and expiry rules for Studio were not stated on the official pricing page
• Avatar video burns 30x voiceover — 48 minutes a year on Studio Starter, not 24 hours
• The API free tier appears to cost the same as or less than Starter at every volume; confirm rate limits before relying on it
• Scale's included characters cost about 11% more each than its own overage rate
• Whether the API plans carry commercial-use rights equivalent to paid Studio tiers was not established from a public page
• No official statement was found on machine-readable watermarking of generated audio
• The CFA complaint of 27 July 2026 remains an allegation with no public regulatory outcome

How it fits against other tools here

Speechify arrives in a category that has lost two listings this year. PlayHT is gone and Resemble AI stopped selling voice AI to new customers in August 2026, which left ElevenLabs, Cartesia and Murf AI as the only live options on the shelf. Speechify fills two slots at once: its API is the new floor on price, well below the position Cartesia has held since the PlayHT correction, and Studio is the cheap browser studio that competes with Murf rather than with an API.

Against Murf specifically, the split is clean. Murf bills Studio, API and Dub as three separately purchased products; Speechify bills Studio and API separately too, but Studio bundles dubbing and avatars into the same credit pool rather than selling them as another line item. Against ElevenLabs, this is a price argument and not much else — ElevenLabs is the reference point for realism and cloning depth, and a 36x list-price gap should make you suspicious enough to audition both. The wider lists are best AI voice tools and the ElevenLabs alternatives shortlist, and everything in the category sits on the AI voice tools page. Because Studio also generates avatar video, it is worth reading alongside HeyGen vs Synthesia if avatars are the actual job.

FAQ

How much does Speechify cost?

It depends which of three products you buy. The reading app lists Premium at $29 a month. Speechify Studio showed annual pricing only in September 2026: $100 a year for 86,400 credits, $300 for 345,600, plus a free plan with 600. The developer API is a third meter at $0, $10, $99 and $499 a month. Nothing on the reading-app pricing page mentions Studio credits, and nothing on the Studio page mentions the reading features.

What does a Speechify Studio credit buy?

One credit per second of voiceover, 3 per second of dubbing, 30 per second of avatar video — a 30x spread against one pool. Studio Starter's 86,400 credits are therefore 24 hours of voiceover, 8 hours of dubbing, or 48 minutes of avatar video for the year. Re-exporting content you already generated is free.

Is the Speechify API cheaper than Speechify Studio?

By roughly ten times for plain narration. Studio Creator is about $0.052 per finished voiceover minute; the API Starter plan is about $6.67 per million characters, or roughly $0.005 per spoken minute at 750 billable characters a minute. The gap is the price of the editor, stock media, cloning and avatars that Studio bundles and the API does not.

Is the API Starter plan worth $10 over the free tier?

On the published numbers, no. Both carry the same $10 per million overage rate, and Starter adds 1M characters for exactly $10, so the two cost the same at every volume from 1.5M characters up and Free is cheaper below that. This assumes the free tier permits metered overage, which the pricing page implies. Check rate limits and support terms before relying on it.

Is Speechify voice cloning consent verified?

On the API, now yes. The Consumer Federation of America asked the FTC and state attorneys general to investigate on 27 July 2026, alleging reliance on "little more than a self-attestation checkbox". A developer changelog entry filed under 13 August 2026 replaced that with a challenge-response flow: Speechify issues a phrase, the speaker is recorded reading it, and Speechify transcribes and checks it against the issued phrase and keeps it as the consent record. CartSignal has not tested the flow and does not assert the complaint caused the change.

Has CartSignal tested Speechify?

No. This is a public-data review built from Speechify's official Studio, reading-app and API pricing pages, its Studio terms and its developer changelog, checked September 2026, plus a dated CFA press release. No hands-on audio quality, latency or cloning testing is claimed.

Source links

• Official Speechify Studio pricing (plans, credit allowances, per-second rates, free re-export)
• Official SpeechifyAI developer pricing (character allowances, agent minutes, overage rates, billing method)
• Official Speechify reading-app pricing (Free and Premium)
• Official Speechify Studio terms of service (ownership, commercial rights, AI disclosure, cloning consent)
• Official developer changelog: voice cloning consent challenge-response flow
• Official Speechify Studio product page (feature list)
• Consumer Federation of America press release, 27 July 2026