Two plans clear 96+ hours a year, so a build past about eight hours a month loses another plan. One plan includes API access, so that final requirement picks your tier for you. ### The limits to check Check four limits, each against the plan that lifts it. Generation hours need the 96+ hour tier, which only two plans reach.  If you need API access, one plan has it. Commercial use needs any of the three paid-rights plans, since Free has none. Concurrency, cloning and seats aren't stated per plan in the sources, so confirm them in writing before you budget. ### API access and concurrency With one plan holding API access, every automation, agent or app call is priced at that tier's rate, whatever your volume. A hobby build at 1,000 minutes a month pays the same entry price as one at 100,000. I can't price the concurrency ceiling. The pricing data I used doesn't publish a parallel-request cap per plan, and a tier that hides this number leaves you finding it in production, as 429 errors. ### Commercial use, voice cloning and downloads If you publish, sell or run a voiceover in a client product, you need commercial rights, and those reach three plans. **Free output is for testing only**. Cloning and download-format limits aren't broken out per plan in the data above, so treat them as unconfirmed. If your product depends on a cloned voice, that's a question for the vendor, not an assumption. ### Murf AI seats, projects and overage charges None of the data I have states seat counts or overage rates. Without an overage rate you can't compute cost per finished minute past the cap, and that gap is where the bill surprises people. ## Murf AI against ElevenLabs, WellSaid and Play.ht pricing I couldn't confirm any of their current prices, so this comparison ranks the limits to check, not a winner on cost per finished minute. Before you build, take from the table which cells you must fill from each pricing page. | | Murf | ElevenLabs | WellSaid | Play.ht | |---|---|---|---|---| | Entry price | Unverified | Unverified | Unverified | Service offline, none current | | Included minutes | Unverified | Unverified | Unverified | Not applicable | | Cost per minute at 10,000 | Unverified | Unverified | Unverified | Not applicable | | API | Offered | Offered | Offered | Not applicable | | Concurrency | Unverified | Unverified | Unverified | Not applicable | | Voices | Unverified | Unverified | Unverified | Not applicable | | Languages | Unverified | Unverified | Unverified | Not applicable | | Rights | Unverified | Unverified | Unverified | Not applicable | | Free tier | Unverified | Unverified | Unverified | Not applicable | "Unverified" means I had no figure from the vendor's own page that I could stand behind. Of all the rows, the API row is the only one I'll commit to, and only as "offered". Whether API access sits on the entry plan or a higher one is exactly the kind of gate this article is about, so check it first. With the service offline, any price you find quoted for Play.ht is stale, so drop it from your shortlist. ### Which Murf AI alternative fits your build Murf's draw is a studio editor for producing finished voiceovers, so it suits you if people, not code, generate most of your audio. If your build is API-first and voice variety matters, ElevenLabs is the one to test. WellSaid targets teams that care about licensing clarity, which makes it the one to read closely on commercial rights. When the pricing page says "contact sales" for concurrency or rights, it hasn't told you your bill. Run one real script of 10,000 minutes through each vendor's calculator and see which limit you hit first.  ## Murf API calls in workflows From a workflow, a Murf call is one authenticated POST that returns audio, plus retry logic for the limit your 10,000-minute test hit first. ### The API call, step by step Five steps, and the order matters: 1. Create the API key. Generate it in your Murf account's API settings. **Keys tie to the plan that issued them**, so a key from the free tier carries the free tier's limits into whatever you build. 2. Store it as a secret. Put it in your workflow tool's credential store or an environment variable, never in the script body. A key pasted into a shared workflow is a key anyone with access can spend against your minutes. 3. POST to the text-to-speech endpoint. Send the text and a voice ID in the JSON body, with the key in the header the docs specify. The voice ID decides which voice you're billed for, so pin it in configuration rather than letting each run choose. 4. Read the audio in the response. Take the audio, or the link to it, from the response and write it to your own storage immediately. A link you never saved is a minute you paid for and may have to buy again. 5. Back off and retry on a rate-limit error. When the API returns a rate-limit response, wait, then retry with a longer delay each time, and stop after a fixed number of attempts. Retrying in a tight loop is how a concurrency ceiling turns into duplicate generations. ### Calling the API through HTTP requests Any workflow tool that offers an HTTP request step can make this call, and that's the honest summary. You won't find the work in a Murf-specific button; you configure the same POST as in the third step, reference the stored secret, and map the response into the next step. This goes wrong at the retries. Check whether your tool's built-in retry applies to rate-limit responses, and whether it waits between attempts. If it doesn't, add the delay yourself, because a workflow that fires many runs at once will hit your concurrency limit before your minute allowance. [Activepieces](https://www.activepieces.com)' published pricing page lists 1 credit per flow run, any number of steps, while Zapier's pricing page bills per task and Make's bills per module, so the backoff, the storage write and the retry cap add nothing to the meter, and a ten-step flow with all of them costs the same as the two-step version without. The Murf side of the bill stays the one this article has been pricing, in minutes. ### How to choose a Murf AI plan Start with measurement. Count the minutes of speech your real scripts produce in a month, not the minutes you hope to sell. Then compare plans by cost per finished minute at that volume, checking which of the four limits (minutes, API access, commercial rights, concurrency) you cross first. Finally, use the free tier to test the limit you suspect, by running the call above in parallel and watching where the errors start. ## What Activepieces does about this Activepieces won't change Murf's limits, but it makes finding yours cheap. Build the five-step call once, using the HTTP step, a stored connection for the key, a delay and a retry cap. Then run it in parallel and see which limit fails first.  The core is open source under the MIT licence, so you can self-host it. The cloud pricing page lists 1 credit per flow run, whatever the step count, so the test flow costs the same as the production one. Log each run's characters and the plan you tested on. That gives you cost per finished minute from your own data. ## Frequently asked questions ### Is Murf AI free to use? Yes, there's a free plan, but treat it as a trial for auditioning voices, not a production tier. Its limits on generation and exports are tight, and its licence terms are narrower than the paid plans', so test your script there and budget as if you'll leave it. ### Can I use Murf voices commercially on every plan? No. Commercial rights are a plan-level term, and the entry-level and free tiers are the likeliest to restrict them. If your output reaches customers, such as a paid course, an ad or a product with synthetic narration, read the licence text for the exact plan before you build. If a pricing grid only hints at rights, that is not enough. If the answer takes a sales call to extract, price in that friction. ### Does Murf charge extra for API access? Plan on it being a separate line item. API usage is metered and governed by its own terms, apart from the generation time you use in the studio editor. **Studio minutes and API minutes are different budgets**, so a plan that looks generous in the editor can look thin once an automation starts calling the API. Work out the cost per finished minute for the path you'll actually run. Confirm in writing whether the API draws from the plan's allowance or bills on top of it. ### Can I cancel or downgrade a Murf plan mid-term? You can stop a plan from renewing, and access normally runs to the end of the period you paid for. Refunds for the unused remainder are a different matter and depend on the billing terms at the time of purchase. Go annual and you commit a year of spend to a volume you may not have measured yet. The discount is real, but that is the risk. **Run a month-to-month plan until your usage settles, then switch if the numbers justify it.** ### Does unused Murf generation time roll over? Assume it doesn't. Allowances on plans like this usually reset each billing period, so unused minutes are spend that bought nothing. The practical consequence is that a lumpy workload, with heavy launch months and quiet ones, costs more per finished minute than a steady one on the same plan. If your volume swings, size the plan to your typical month rather than your peak, and handle spikes another way. Check the plan's terms for any carry-over exception before you rely on either assumption. ## Related reading - [Claude Sonnet 5.5 API Pricing for Developers (2026)](https://www.activepieces.com/blog/claude-sonnet-55-api-pricing-for-developers-2026) - [Firebase Studio Pricing 2026 Explained for Builders](https://www.activepieces.com/blog/firebase-studio-pricing-2026-explained-for-builders) - [Best Tools for Citizen Developers in 2026](https://www.activepieces.com/blog/tools-for-citizen-developers-in-2024) ## References - [Murf](https://searchsagar.com/pricing/murf-pricing/) - [Jellypod](https://www.jellypod.com/blog/murf-ai-pricing) - [Price Per Prompt](https://www.priceperprompt.com)Limits, not price, force the upgrade: add commercial rights, 96+ generation hours a year and API access to one build, and only one Murf plan still qualifies.