Best AI Dubbing Software
Published September 2, 2026 · Every claim below was read from the vendor’s own pages on September 2, 2026. No listening test has been run, so nothing here says how any dub sounds — every statement about voice preservation, lip-sync, timing or accuracy is a quoted vendor claim, reported as a claim.
The short answer
This market is far smaller than it looks. Six of the ten providers we track publish no dubbing product at all. Amazon Polly, Google Cloud, OpenAI, MiniMax, Cartesia and Resemble AI sell nothing you could dub a video with.
One of those six advertises a dubbing product it does not have. Cartesia publishes a page headlined “Fastest multilingual AI video dubbing” describing an upload step that exists in none of its documentation.
That leaves four: ElevenLabs, Azure, Murf AI and Speechify. Only ElevenLabs publishes a dubbing price you can read on a page. Azure’s prices exist in the page’s markup but render to you as a dash.
And three of the four contradict themselves on how many languages they dub. Murf publishes five different figures for one product on its own pages on the same day. This is the single least reliably documented capability in the market.
Basis: vendor documentation only, checked September 2, 2026 — no accounts created, no dubs generated, no audio evaluated (how we verify).
On this page
Who actually offers dubbing
Before any ranking, the question nobody else seems to ask: which of these vendors sells dubbing at all? We checked all ten. Four do.
| Provider | Dubbing product? | What we found |
|---|---|---|
| ElevenLabs | Yes | “Dubbing” — two models, two surfaces, a documented API, and a sales-gated human-verified tier |
| Azure Speech | Yes | “Video translation” — no separately branded dubbing SKU, but Microsoft uses “dub” for the function |
| Murf AI | Yes | “Murf Dub” plus a “Murf Dub Automation API”, on a separate app and a separate API key |
| Speechify | Yes | One capability under four different vendor names across four of its own pages |
| Cartesia | No — but it advertises one | A marketing page headlined “Fastest multilingual AI video dubbing” with no product documented anywhere |
| Amazon Polly | No | AWS names, sells and documents no dubbing or video-localization product; Polly is documented solely as text-to-speech |
| Google Cloud | No | The one dubbing URL in Google’s entire sitemap returns HTTP 404 and sits under a private path |
| OpenAI | No | The word “dubbing” appeared on no official page. Realtime translation exists and is not presented as dubbing |
| MiniMax | No | Neither documentation index returns any match for dubbing, translation, localization, lip-sync, subtitles or captions |
| Resemble AI | No | No dubbing product, and as of August 13, 2026 no commercial voice product at all — see the correction |
Two of those negatives deserve their method stated, because “we could not find it” and “it does not exist” are different claims. For Google Cloud we enumerated its sitemap — 180 sub-sitemaps — and searched every URL for “dub”. Exactly one matched, it sits under a /docs/private/ path, and it returns HTTP 404 on both of Google’s documentation hosts. For MiniMax we read the machine-readable documentation indexes for both its international and mainland platforms, 25,188 and 26,044 bytes, and neither returns a match for dubbing, translation, localization, lip-sync, subtitles or captions.
The page with no product behind it
This is the sharpest finding in the research, and it is the reason we check vendor claims against vendor documentation rather than against vendor marketing.
Cartesia publishes a page at cartesia.ai/product/ai-dubbing headlined “Fastest multilingual AI video dubbing”. It contains a numbered how-to, step two of which reads verbatim: “Upload your video content and select the desired languages for dubbing using Cartesia’s intuitive platform.”
No such upload exists in any Cartesia documentation. There is no dubbing endpoint, no video ingestion, no dubbing parameter and no dubbing entry in its pricing. Cartesia’s documented products are text-to-speech, speech-to-text and voice agents.
The same page carries two further claims its own documentation contradicts: a “time-to-first-audio of just 40ms” against the documentation’s “about 90ms” for Sonic 3.5, and cloning “from a 3 second clip” against the documentation’s “as little as 10 seconds of audio”. When a marketing page disagrees with the documentation three times on one page, the documentation is what we report. Our Cartesia review covers what the company actually ships.
ElevenLabs
The only one of the four that publishes a dubbing price you can read. Ranked first on documented capability, price transparency and licensing clarity — not on output quality, which we have not tested.
What it is. Two named models, Dubbing v1 and Dubbing v2, across two surfaces: “Automatic Dubbing” (“Dub content from one language into another with a few clicks”) and “Dubbing Studio” (“Granular control over your dubs, including transcript editing, speaker reassignment, and per-clip regeneration”). Dubbing Studio runs only on v1, and the vendor states it “is in maintenance mode and receives critical bug fixes only”. A human tier exists behind sales: “Human-verified dubs via ElevenLabs Productions”.
What it costs. Published per minute, on a page, per tier: Dubbing v1 at $0.33 per minute with a watermark and $0.50 without, identical across all six tiers; Dubbing v2 at $2.20 per minute, also identical across tiers. The metering is stated plainly: “Dubbing is billed per source audio minute.”
Who can use it. Not gated. Verbatim: “Dubbing is available on all our plans, including the free plan. Dubs generated on free plans are automatically watermarked, with no option to remove this.” The practical gate is allowances rather than access — v2 minutes and unwatermarked v1 minutes are both marked unavailable on Free and Pay-as-you-go. Concurrency is documented: “All self-serve plans… allow up to 3 concurrent dubbing jobs, and Enterprise plans default to 10.”
Why it ranks first. It is the only one of the four publishing both an enumerated per-model language table and a documented API parameter surface for dubbing mechanics — num_speakers (“Set to 0 to automatically detect the number of speakers”), drop_background_audio, disable_voice_cloning, a documented limit of “up to 32 unique speakers per file”, and a numeric cloning-strength control.
What is wrong with it, published alongside the rank. Dubbing v2 is self-labelled “currently in alpha”, and “Dubs created using the v2 model are completely automatic without any option to edit the content”. Lip-sync is explicitly not offered — verbatim: “At the moment, ElevenLabs does not offer lip syncing as part of Dubbing.” Real-time dubbing is “not currently available”. And the vendor contradicts itself three times: on v1 and v2 language counts, on upload limits (1 GB in one place, 2 GB in another), and on whether the $2.20 v2 rate is real — its own pricing FAQ omits it while the table on the same page shows it.
Microsoft Azure
Prices published in the page’s markup and rendered to you as a dash. Second on documented capability and enterprise footing; the price transparency problem is a rendering failure rather than a refusal to publish.
What it is. “Video translation”, described as a feature “that enables you to seamlessly translate and generate videos in multiple languages automatically”. There is no separately branded dubbing SKU, but Microsoft uses the verb for the function: its transparency note says the service “will automatically extract dialogue audio, transcribe, translate and dub the content with prebuilt or personal voice to the target language”.
What it costs, and the finding that matters. The Azure Speech pricing page has a “Video Translation” section with three line items billed, in Microsoft’s own unit, per hour. All three render to a human visitor as the literal string $-. The figures exist in the page’s own markup, in the rows’ data-amount attributes: 5.0, 15.0 and 20.0 per hour. So Azure does publish per-hour dubbing pricing. You simply cannot see it by looking at the page, which is the same defect that makes its text-to-speech rates unreadable — the pricing page renders $- 142 times, as our pricing comparison records.
We publish those three numbers as what the vendor’s own page carries, and we do not tell you which line item is which beyond what the page states, because the rendering failure extends to the labels a buyer would use to match rate to service.
Murf AI
The best-documented capability of the four, and no published dollar price at all. Third: capability is real and enumerated; the price is behind a login.
What it is. “Murf Dub”, on its own application at dub.murf.ai, plus a “Murf Dub Automation API”. The API needs its own key — verbatim: “API Key for Murf Dub Automation API is different from the one used for other Murf services (like TTS).”
What it costs. No dollar figure for dubbing is published on any public Murf page. What is published is a free allowance: “100 minutes of free dubbing after sign up”. The unit is credits — “Each dubbing consumes credits based on the minutes and number of languages” — and no credit-to-currency rate is published. Its API pricing page prices exactly four line items, and dubbing is not one of them.
One number on that page is not what it looks like. Murf’s dubbing page displays “$75-$100 per minute”. That figure sits in a column headed “Traditional Dubbing” and is Murf’s characterization of human studio dubbing, not its own price. It is quoted here because it is the most prominent dollar figure on the page and the easiest thing on this market to misread.
Why it still ranks third rather than lower. Its documentation is the most substantive of the four. It is the only vendor publishing per-mode source and destination language tables with locale codes, a documented tier-limits table, an auto-detect source locale, two API modes, and a human QA add-on. It also states an uptime position: “We typically offer 99% uptime.” Free-tier output appears to be watermarked — “No watermarks” is a paid-plan feature string in Murf’s own application code and is stated on no public page, so we report it as the weak evidence it is.
Speechify
Thinnest documentation of the four, and the only one whose commercial rights over dubbed output are genuinely undetermined. Fourth on all three axes.
What it is. One capability under four vendor names across four of its own pages: “Dubbing Studio”, “AI Dubbing”, “AI Dubbing Online For Video Localization” and “Video Dubbing Editor”. Any single name you attribute to Speechify will be wrong on three of its own pages.
What it costs. No per-minute, per-video or per-dub price is published anywhere. Dubbing is metered in Studio credits and only the burn rate is documented: “Dubbing → 3 credits per second”, against “Voiceover → 1 credit per second”. Plans are $0 for 600 credits, $100/year for 86,400, and $300/year for 345,600. No credit-to-currency rate is published, so three credits per second cannot be converted into money, and we have not converted it.
Who can use it. Not gated: “Dubbing Studio” is a bullet on the free plan card. The limiter is credits — 600 free credits at 3 credits per second is 200 seconds. What is gated alongside dubbing is more consequential than dubbing itself: the free plan card reads “No Voice Cloning” and “No commercial usage rights”.
Why it ranks last. Speaker separation, diarization and multi-speaker handling are documented nowhere — a case-insensitive search of the full served HTML of its three dubbing pages for “speaker”, “diariz”, “separat”, “multiple voices” and “original voice” returned nothing beyond the single phrase “Match the speaker’s voice, intonation, and speed.” There is no API, no parameter surface, and no in-app documentation readable without an account.
What dubbing costs
| Provider | Unit | Published rate | Free allowance |
|---|---|---|---|
| ElevenLabs | Per source audio minute | v1 $0.33 watermarked, $0.50 unwatermarked; v2 $2.20 | 2.53 v1 watermarked minutes/month |
| Azure | Per hour | 5.0 / 15.0 / 20.0 — in the page markup only; renders as $- | Not stated for video translation |
| Murf AI | Credits, by minutes × languages | None published — no credit-to-currency rate | 100 minutes after sign-up |
| Speechify | Credits per second | 3 credits/second — no credit-to-currency rate published | 600 credits, i.e. 200 seconds |
Only one row in that table lets you answer “what will it cost me to dub a ten-minute video?” without an account. On ElevenLabs v1 with a watermark that is $3.30, and on v2 it is $22.00, both by simple multiplication of a published per-minute rate. For Murf and Speechify the question is unanswerable from public pages, and for Azure it is answerable only by reading an HTML attribute.
Note also that ElevenLabs meters source minutes: dubbing one ten-minute video into six languages bills ten minutes, not sixty, on that vendor’s stated unit. Murf’s credit consumption explicitly scales with “the minutes and number of languages”. Those are opposite billing models and nobody puts them side by side, so we have.
The language-count problem
Three of the four contradict themselves on how many languages they dub — not between vendors, but within each vendor’s own pages on the same day. Azure is the exception only because it publishes no dubbing language count at all, which is its own kind of gap.
| Provider | Figures published on its own pages | What its enumerated list actually contains |
|---|---|---|
| ElevenLabs | Docs say “90+”; pricing cards say 29 (v1) and 92 (v2) | We counted the docs’ own tables: 86 rows for v1, 94 for v2 |
| Murf AI | Five figures: 20+ (navigation), 28 (API docs), 28+ (subhead), 30+ (structured data), 40+ (page title and hero) | 22 source locales; 26 automated destinations; 19 destinations for QA dubbing |
| Speechify | “Dub Content in 60+ Languages” | 52 unique names in the list rendered directly beneath that heading |
| Azure | No dubbing-specific language count found | No enumerated video-translation language list found |
Murf is the extreme case: five different numbers for one product, on one vendor’s pages, on one day, while its own enumerated tables — the only countable thing it publishes — list 22 sources and 26 destinations. The “40+” in its page title could not be reconciled with any published list.
Speechify’s is the most direct: a heading claiming “60+” sits immediately above a list of 52. You do not need our verification to check that one; you need to count the list under the heading.
What we do with this. We publish no single language number for any of these products. Where a vendor enumerates a list, we report the count we made ourselves and say we counted it. Where it only asserts a figure, we report the assertion as an assertion. Reconciling a vendor’s contradictory claims into one tidy number would be inventing a fact none of them published.
What nobody publishes: language pairs
Dubbing is a directional operation. It goes from a source language to a target language, and whether a given pair works is the question you actually need answered before committing to a vendor.
Not one of the four publishes a source-to-target pair matrix. ElevenLabs documents source and target tags independently — “The API accepts a BCP-47 language tag in the source_language and target_language parameters” — and dialect support is documented for exactly six languages on v2. Speechify documents target languages only. Murf comes closest, publishing separate source and destination tables, but stops short of stating which combinations are supported.
So if your requirement is Korean into Brazilian Portuguese, no vendor page in this market will confirm it works before you sign up. That gap is the same across all four, and it is the most consequential thing missing from this entire category.
Lip-sync: who claims it
Lip-sync is the feature most likely to decide whether AI dubbing suits a video, and the four vendors are in three different positions on it.
- ElevenLabs says no, plainly. Verbatim: “At the moment, ElevenLabs does not offer lip syncing as part of Dubbing.” That is the clearest statement any vendor makes on this page, and it is a refusal.
- Murf says both. A dedicated feature card on its dubbing page reads “Hyper Accurate Lip-Sync (Coming Soon)”, while the same page’s subhead and FAQ describe lip-sync as something the product does. The only unambiguous statement on that page is the “(Coming Soon)” label, so that is what we report.
- Speechify claims it in marketing prose only. Verbatim: “With advanced AI dubbing and precise lip-syncing capabilities…” and “…perfect pronunciation and lip-synching”. No control, no flag, no parameter and no documented limitation accompanies either sentence.
- Azure — we found no lip-sync claim or denial for video translation either way.
A vendor that says it does not do something is more useful than a vendor that says it does something perfectly and documents nothing about it. That is a statement about documentation quality, not about output, which we have not evaluated.
Who owns a dubbed video
None of the four publishes a dubbing-specific licensing clause. You are governed by general terms written for other products, and in one case those terms and the pricing page point in opposite directions.
Speechify is the genuinely unresolved one. Its Terms clause 4.4 grants commercial use only for “Speechify Voice Over Studio AI voice overs”. Its pricing page sells “Commercial usage rights” as a paid feature listed alongside “Dubbing Studio”. No page reconciles the two, and the commercial status of dubbed output is undetermined from official sources. If you are dubbing anything you intend to sell, that is the sentence on this page to act on.
Murf’s governing clause places the burden on you. Its dubbing-relevant clause is §4.2 on voice cloning, verbatim: “The submission of a third party’s unauthorized voice recordings as training audio is prohibited, and you agree that you will not submit any recording to Murf that contains the voice of any speakers that have not explicitly consented in writing…” Dubbing an interview means warranting written consent from everyone audible in it.
ElevenLabs has no dubbing-specific clause either, but its governing clauses are numbered and quotable, which is more than the other two offer. The full licensing position for all ten providers is in our guide to can you use AI voices commercially.
Frequently asked questions
What is the best AI dubbing software?
Of the four vendors that actually sell it, ElevenLabs is the only one that publishes a price you can read, an enumerated language table and a documented API parameter surface. That is a documentation judgment, not a quality one — we have run no listening test, so we cannot tell you which produces the better dub.
Can I dub a video for free?
Yes, with three of the four. Murf offers 100 free minutes after sign-up, ElevenLabs includes about 2.5 watermarked minutes a month on its free plan, and Speechify’s 600 free credits buy 200 seconds. ElevenLabs and Murf free output appears to be watermarked, and Speechify’s free plan carries “No commercial usage rights”.
How much does it cost to dub a ten-minute video?
Only one vendor lets you work that out from public pages. On ElevenLabs, $3.30 on v1 with a watermark or $22.00 on v2, from a published per-minute rate. For Murf and Speechify no credit-to-currency rate is published, so the question cannot be answered without an account.
Does dubbing into more languages cost more?
It depends on the vendor, and they are opposite. ElevenLabs states dubbing is “billed per source audio minute”, so the source length is what meters. Murf states credits are consumed “based on the minutes and number of languages”, so each additional target language adds cost.
Which AI dubbing tool does lip-sync?
ElevenLabs states it does not. Murf labels its lip-sync feature “Coming Soon” on the same page that describes it as working. Speechify claims it in marketing prose with nothing documented behind it. No vendor here publishes a lip-sync control, limitation or specification.
Can I check whether my language pair is supported before signing up?
No. None of the four publishes a source-to-target pair matrix. Three publish target lists and one publishes separate source and destination tables, but none states which combinations work.
Why do so few of these providers offer dubbing?
Dubbing is a different product from speech synthesis — it needs transcription, translation, speaker handling and video processing on top of a voice model. Six of the ten sell the voice model and nothing built on it. That is a market structure fact, not a criticism.
Can I trust a vendor page that advertises dubbing?
Check the documentation, not the landing page. One vendor in this comparison publishes a dubbing product page, complete with an upload step, for a capability that appears nowhere in its documentation, pricing or API.
Sources and what we could not verify
Every claim on this page was read from the vendor’s own pages on September 2, 2026, and every quotation is reproduced verbatim. Negative findings were established by enumerating vendor sitemaps and machine-readable documentation indexes rather than by search, and the method is stated above for the two cases where it matters most. Per-provider detail is in the reviews linked throughout, and the verification method is on the methodology page.
What we could not verify
Honesty about gaps beats a page that looks complete. Genuinely unresolved as of September 2, 2026:
- How any of these dubs sound. No listening test has been run. Every statement about voice preservation, timing, emotion, accuracy or lip-sync on this page is a quoted vendor claim.
- Any single language count, for any of the four. Every vendor contradicts itself; we publish the conflicts rather than a number.
- Whether any specific source-to-target pair works. No vendor publishes a pair matrix.
- What Murf and Speechify dubbing costs in money. Both meter in credits and neither publishes a credit-to-currency rate.
- Which Azure video-translation rate applies to which line item. The figures exist only in the page’s markup, and the rendering failure extends to the labels.
- Whether ElevenLabs Free and Pay-as-you-go users can buy Dubbing v2 minutes. A v2 per-minute price is listed for that tier while its included-minutes cell is marked unavailable, and no page resolves it.
- Whether Speechify dubbed output may be used commercially. Its terms and its pricing page point in opposite directions and nothing reconciles them.
- Whether Murf free-tier dubs are watermarked. Implied by a paid-plan feature string in Murf’s own application code; stated on no public page.
- ElevenLabs upload limits. One page says 1 GB, another says 2 GB.
- Enterprise dubbing pricing anywhere. Not published by any of the four.
Change log
September 2, 2026 — First publication. All claims read from official vendor pages on September 2, 2026 and dated accordingly.
Published September 2, 2026 · This page shows no audio and reports no listening results, because our benchmark has not run and does not yet cover dubbing. Vendor pages in this category change frequently and several carry no version or date; corrections are recorded with their date on the corrections page. 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.