Which AI voice generator should you pay for? What an hour of voiceover costs on each

Short answer: for voiceovers you publish, pay for ElevenLabs Creator ($22 a month, about 121 minutes of speech) if one to two hours a month covers you, or Murf Business ($792 a year, 96 hours a year) if you make a lot of corporate or training audio. By our arithmetic from each vendor's own pricing page, read on November 3, 2026, an hour of narration costs about $8 to $11 on the app plans made for creators and $0.60 to $1.80 on the developer APIs from Murf, Google, Microsoft Azure and OpenAI. The APIs are cheaper by roughly ten times, but you need a script or a tool built on them, and neither ElevenLabs' nor Murf's free plan lets you use the audio commercially.

How we got an hour: ElevenLabs' own plan table equates 10,000 credits (1 credit per character on its main models) with "~10" minutes of speech, so about 1,000 characters is a minute and 60,000 characters is an hour. We used that for every per-character price. A slower read packs fewer characters into an hour, so your real cost can be a little lower. App-plan hours assume you use every included minute.

How much does an hour of AI voiceover cost on each tool?

About $8 to $11 on the app plans and under $2 on the APIs. Here is each price as the vendor published it on November 3, 2026, and what that means per hour.

• ElevenLabs. Its pricing FAQ lists "Free $0 (10,000 credits); Starter $6 (30,000 credits); Creator $22 (121,000 credits, $11 for the first month); Pro $99 (600,000 credits)". For its V2 Multilingual models, "1 text character equals 1 credit". Unused credits roll over for up to two months on a paid plan, and yearly billing is "monthly price × 10". On Nov 3 the page also introduced Eleven v4, which ElevenLabs calls "our most emotive model yet".
• Murf. On the Studio tab, Creator is "$19 / month" billed as "$228 Billed Annually" with "24 hrs/Year of Voice Generation"; Business is "$66 / month", "$792 Billed Annually", "96 hrs/Year". The API tab lists text to speech at "$0.01 / 1000 characters" (Falcon 2.1, built for real-time voice agents) and "$0.03 / 1000 characters" for its studio-quality model, plus "$10 FREE for API calls".
• Google Cloud. Chirp 3: HD voices are free for "0 to 1 million characters" a month, then "US$30 per 1 million characters". Gemini TTS is billed per audio token, and the page notes "Audio tokens correspond to 25 tokens per second of audio".
• Microsoft Azure. The pricing page gives "0.5 million characters free per month" for Neural voices on the free tier. Azure's own Retail Prices API lists Neural text to speech at $15 per 1M characters and Neural HD at $22 per 1M in East US.
• OpenAI. tts-1 is "$15.00 / 1M characters" and tts-1-hd "$30.00 / 1M characters", but see the shutdown date below.

Which AI voice generator should you pay for?

Pick by the job and by how many hours a month you make. Quality is something you judge with your own ears: both ElevenLabs and Murf have free plans, so run the same paragraph of your real script through each before you pay.

• Faceless YouTube narration, one to two hours a month: ElevenLabs Creator. About 121 minutes a month, a commercial license and professional voice cloning, and unused credits roll over for two months, so a slow month isn't wasted. If you plan a faceless YouTube channel, this is the plan most people start on.
• Ads, explainers and corporate or training audio, several hours a month: Murf Business. It's the cheapest app hour we found ($8.25), its plans list Canva and PowerPoint integration, but you pay $792 for the year up front.
• Audiobooks and other long-form: the hours add up, so price the whole book. A 10-hour book is about 600,000 characters, which is one month of ElevenLabs Pro ($99) or about 10 of Murf Creator's 24 yearly hours. Audition voices on a full chapter, not a sentence: long reads expose pacing problems.
• Dubbing a video into other languages: ElevenLabs prices dubbing at 2,000 credits a minute with a watermark and 3,000 without. An hour dubbed without a watermark is 180,000 credits, more than Creator's 121,000, so plan on Pro. Have a native speaker check the result before you publish.
• Bulk or automated audio, and you can run code: an API. Google's 1 million free characters a month is about 16 hours of Chirp 3: HD speech by our conversion; Azure gives 0.5 million free. After that, $0.60 to $1.80 an hour.
• You already make images and video in Higgsfield: it has text to speech too (see below), so you can try voices without another subscription.

Can I use a free AI voice plan for YouTube or client work?

Not on ElevenLabs or Murf. ElevenLabs' plan table shows commercial use crossed out on Free, and the commercial license starts on Starter ($6 a month, about 30 minutes). Murf's Free plan says "No Commercial Rights" and allows "10 minutes of Voice Generation". The cloud APIs are different: Google's and Azure's free monthly allowances are simply the first characters of a paid service, with no separate "personal use only" plan. Read each vendor's terms for what you publish.

What happened to Play.ht?

It's gone as an option. Meta acquired PlayAI, the company behind Play.ht, in July 2025, a deal "confirmed by a Meta spokesperson" per CXM World. When we tried play.ht on November 3, 2026, the site didn't load at all. If an older guide (including an earlier version of this one) sent you there, pick from the tools above instead. Don't pay a site that only looks like it.

Is OpenAI's text to speech still worth building on?

Not for anything new. OpenAI's text-to-speech guide says "The tts-1, tts-1-hd, and gpt-4o-mini-tts models are deprecated and will shut down on January 6, 2027." OpenAI points developers to its Realtime API instead, which is a different workflow. Any tool you buy that runs on those models will have to switch before then.

Does Higgsfield do AI voiceovers?

Yes. Higgsfield's Audio page has Text to Speech, Voice Change and Translate. Its text to speech runs on a model called Seed Audio 1.0, lets you pick a stock voice or clone one, and has a voice-details box of up to 500 characters for describing how it should sound. We found no per-hour audio price on Higgsfield's pricing page; the credit cost shows in the app before you generate. It makes sense when voice is one step in a video you're already making there, next to the AI video generators and a consistent AI influencer. For hours of narration alone, the dedicated tools above are priced more clearly.

What gives an AI voice away, and how do you fix it?

Wrong-word emphasis, flat emotional beats, mispronounced names and acronyms, and breaths in odd places. Most of it is fixed in the script, not the model: punctuate the way you want it read, spell tricky names the way they sound, break long paragraphs into short lines so pacing resets, and listen to every take once before you publish. Budget for retries too. If you regenerate a third of your lines, an hour of finished audio costs a third more than the table above.

Do I have to disclose an AI voice, and can I clone someone's voice?

Disclose when the result looks or sounds real enough to mislead, and only clone a voice you have permission to use. YouTube's help page says "we require creators to disclose content that is generated or meaningfully altered with AI when it appears realistic", and lists "Cloning one's own voice to create voice overs or dubs" among the things you don't need to disclose. "Making it appear as if someone gave advice that they did not actually give" must be disclosed. If you build an app on an API, OpenAI's usage policies require "a clear disclosure to end users that the TTS voice they are hearing is AI-generated and not a human voice." For a brand voice, use a stock voice or a voice you own.

More ways people turn these tools into work are in the AI Money hub.

Pairing a voice with AI footage? The best AI video generator for a faceless YouTube channel explains why to turn the video model's own sound off under a voiceover, and what it saves.