Google Cloud Text-to-Speech vs Murf AI 2026

Published September 1, 2026 · This is a documentation-based comparison. Every figure below was read from the two vendors’ own pricing sources, documentation and terms on September 1, 2026. No listening test has been run, so nothing here compares how the two providers sound.

The short answer

A sourcing note that belongs beside every Murf price above. Murf’s pricing pages render their figures with JavaScript and we could not read prices from the rendered page. The plan prices here were parsed from Murf’s own pricing data endpoint on September 1, 2026, which returns bare integers with a currency field and a billing-frequency key. They are Murf’s own published values, but not what a visitor sees painted on the page. We do not publish Murf’s annual figures, because those integers carry no unit label.

On this page

Winners by category

Every winner below is a documentation-based judgment from published prices, terms and feature documentation. None is a test result, and the categories that need listening have no winner.

Price and value

Expressed in the same unit, Google’s Standard and WaveNet voices cost $4 per million characters. Murf’s published API rates work out to $10 per million on Falcon or $30 per million on Gen 2 — if you accept the page that splits by model. On the page that does not, the flat rate is also $30 per million.

So Google is cheaper by somewhere between two and a half and seven and a half times, depending on which Murf page governs and which model you use. That range exists because Murf publishes two answers, not because we could not decide.

Murf’s API rate has two official answers

Both live on Murf help pages in the same section on September 1, 2026. One states: “Pay as you go: With our pay-as-you-go plan, you can purchase 1000 characters at $0.03 (minimum purchase amount is $2).” The other states a $10 minimum and splits by model: “Gen 2 model: $0.03/1000 characters. Falcon model: $0.01/1000 characters.” A third Murf page supports the Falcon figure independently.

We publish both and label the conflict. The minimum purchase is contradictory too — $2 on one page, $10 on the other.

Google’s premium models change the comparison

The $4 figure is Google’s older Standard and WaveNet families. Neural2 and Polyglot list at $16 per million, Chirp 3: HD at $30, and Studio voices at $160 per million characters. At the Studio tier Google is the more expensive of the two by a wide margin, so compare like with like for the voice family you actually intend to use.

Google’s Gemini-TTS models are billed in tokens rather than characters — for instance $0.50 per million input text tokens and $10.00 per million output audio tokens on Gemini 2.5 Flash TTS, with the footnote that “Audio tokens correspond to 25 tokens per second of audio”. Those cannot be placed on the same axis as a per-character rate without measurement, and we do not attempt it.

Free tiers

Google gives four million characters a month on Standard and WaveNet, and one million on the premium families, as ordinary metered usage that happens to cost nothing. The procedural catch: “You must enable billing to use Text-to-Speech, and will be automatically charged if your usage exceeds the number of free characters allowed per month.” A card comes before the first free character, and the Gemini-TTS models are excluded from the free tier entirely.

Murf’s answer depends on the product. The Studio trial produces nothing you can keep — “The ability to download and export projects is only available with a paid plan.” The API documents a separate free trial with a real character allowance that does produce usable output.

What this means for you: for evaluating output quality at volume, Google’s free tier is in a different league. For evaluating a workflow, Murf’s Studio trial lets you use the editor but not keep the result.

Commercial use, licensing, ownership

Both vendors leave a gap here, and they leave different ones.

Google never addresses commercial use. Neither the Cloud Platform Terms of Service nor the Service Specific Terms contains a clause permitting or prohibiting commercial use of synthesized audio. What exists is a general intellectual-property provision under which the customer retains rights in Customer Data. That is normal for infrastructure, and it means your rights are inferred from general terms rather than granted for this product.

Murf grants commercial use explicitly, with restrictions attached. Clause 5.1 of its Terms, in a document dated December 19, 2024, begins: “You can use Murf created voices for commercial purposes. You can also make available to third parties who intend to use Murf created voices for commercial purposes.” It continues: “You however agree that Commercial Rights (defined below) are not resell rights and you cannot resell or offer to resell Murf’s offerings (including the library of voices available on Murf Studio) to any third-party. You further agree not to use Murf created voices for training any AI model or to synthesize Murf created voices in any way or for any purpose.”

But Murf never says you own the audio. Clause 3.1 grants ownership of “the content and information that you submit to the Services” — the input. We searched the Terms in full and found no clause assigning ownership of generated output. A commercial permission is not title.

What this means for you: if your client contract requires you to warrant that you own the audio you deliver, neither vendor hands you that warranty on a plate. Google’s general terms come closer by stating you retain rights in your data; Murf gives you permission to sell work it never says is yours.

One further asymmetry for data-sensitive work: Murf’s no-training covenant — that your text and audio never feed model training — is documented as an Enterprise commitment rather than a property of the ordinary paid plans.

Voice cloning

Neither of these providers will let you clone a voice today without talking to someone.

Google’s Instant Custom Voice is allow-listed: “Access to Instant Custom Voice is restricted to allow-listed users. To request access, contact a member of the sales team.” It carries a published rate of $60 per million characters, so the pricing is transparent even though the access is not open.

Murf’s cloning is an enterprise engagement requiring studio-quality recordings, a multi-week turnaround and explicit written consent. There is no self-serve tier at any price.

If self-serve cloning is what you came for, ElevenLabs documents it from $6 a month and Cartesia from $5 (both checked September 1, 2026). The ElevenLabs versus Murf comparison covers that contrast directly.

Workflow and API

Google ships no editor. It is a synthesis endpoint inside a cloud platform, and everything around it — project structure, revisions, multi-speaker scripts — is yours to build. Murf ships the opposite: a timeline studio where the API is the secondary product.

One dated item belongs to Murf and is worth checking before you build on its API: the legacy Gen2 streaming model was documented for deprecation on 16 August 2026, a date that has now passed. As re-checked on September 1, 2026, Murf’s documentation describes it as deprecated in the present tense while its OpenAPI specification still lists gen2 as an accepted streaming value. Murf publishes no removal notice, and we do not claim one.

On Google’s side, the corresponding planning note is the unit split: per-character models let you forecast from script length; the Gemini-TTS family, billed in tokens, does not.

Voice quality, pronunciation and naturalness

We have not compared how these two providers sound.

Our standardized listening benchmark is designed but not running. Until it does, this page carries no naturalness verdict, no pronunciation comparison and no scores for either provider. Both vendors publish quality claims of their own; those belong to them, and we do not repeat them as findings.

Choose Google if · Choose Murf if

Choose Google Cloud Text-to-Speech if:

• Volume is your problem. $4 per million characters on Standard and WaveNet, with no plan to outgrow.
• You want a genuinely usable free allowance: four million characters a month, recurring.
• You are already on Google Cloud and want regional control and IAM for free in engineering terms.
• You are building a pipeline, not producing a voiceover.

Choose Murf AI if:

• Your work is scripted voiceover with revisions. The timeline studio is the reason to be here, and Google has no equivalent at any price.
• You need a commercial grant you can quote to a client — carrying its two restrictions.
• You want a product a non-developer can operate. Google’s requires a billing account and code before the first file.
• Your volumes are modest, where the per-character gap costs a few dollars and the workflow gap costs days.

Sources and what we could not verify

Every figure on this page was read from the two vendors’ official sources on September 1, 2026. Full detail is in the individual reviews: Google Cloud Text-to-Speech review and Murf AI review. How we verify anything is described on the methodology page.

What we could not verify

Honesty about gaps beats a page that looks complete. Genuinely unresolved as of September 1, 2026:

• How either provider sounds. No listening test has been run.
• Murf’s actual API rate and minimum purchase. Two official pages disagree on both; we publish both.
• Murf’s prices as a visitor sees them. Its pricing pages render figures with JavaScript; ours come from Murf’s own data endpoint. Annual figures are not published here.
• Who owns Murf’s generated audio. No clause in its terms assigns it.
• Whether Google permits commercial use. Neither governing document says so either way; we report the silence rather than fill it.
• Any per-character comparison against Google’s Gemini-TTS models. They bill in tokens and no conversion is published.
• Whether Murf’s Gen2 streaming model still works. Documented as deprecated in the present tense, still listed as an accepted streaming value, with no removal notice.

Change log

September 1, 2026 — First publication as a documentation-based comparison. All figures read from official vendor sources on this date.

Published September 1, 2026 · When our audio benchmark runs, this page gains a head-to-head listening test using the same prompts and documented conditions for both providers, and the naturalness and pronunciation rows gain real winners (methodology). Independence note: this page contains no affiliate links, and no vendor paid for placement or influenced the order — see how we make money and our editorial policy.