Vaani
Vaani is an AI dubbing platform that translates video into more than forty languages using the original speaker's voice, cloned straight from the source footage, with scene-aware translation and an optional, face-aware lip-sync pass.
What is Vaani?
Vaani is an AI dubbing platform built around one distinguishing choice: it clones the speaker's voice from the source video itself. Where most dubbing tools ask you to record separate voice samples or pick a narrator from a stock library, Vaani needs only a few seconds of someone talking on camera. The result is meant to sound like the same person speaking another language, rather than a synthetic narrator laid over your footage.
Language coverage leans heavily towards South Asia. Thirteen Indian languages are supported — Hindi, Malayalam, Tamil, Kannada, Bengali, Gujarati, Marathi, Punjabi, Telugu, Odia, Assamese, Urdu and Nepali — alongside thirty global ones, from Spanish and Mandarin to Pashto, Swahili and Hausa. Translation is described as scene-aware: the model reads the visual context of each shot next to the transcript, so idioms, cultural references and tone follow what is happening on screen instead of being rendered sentence by sentence in isolation.
Lip-sync is an optional layer, and its billing is the notable part. Vaani counts only frames where a face is on camera and speech overlaps, so B-roll, cutaways and silent reaction shots pass through unbilled. Two engines are offered: a Normal pass on Replicate's sync/lipsync-2 for cost-sensitive work, and a Pro pass on sync.so's sync-3 for sharper edges, profile angles and fast-cut footage.
Work happens in one of two surfaces. Studio is a DAW-style timeline for the single project that has to be right, with per-segment editing, per-speaker voice settings, a lip-sync layer and manual override on every line. Glot is a node-based board for volume: drop many files, fan them out across target markets, watch the renders run in parallel, download a zip. Behind both sits the same four-stage pipeline — vocal isolation on dedicated GPUs, transcription with word-level timestamps and speaker diarisation, transcreation, then a loudness-normalised broadcast mix.
A REST API with HMAC-signed webhooks exposes the same engine from the Studio plan upwards. The pipeline does not change with your tier: only the cost per action moves.
What it does
- Dub a video into 40+ languages using the original speaker's cloned voice
- Clone a voice from the source footage itself, with no separate recording or training step
- Translate with scene-aware context, so idioms and tone follow what is on screen
- Transcribe with speaker diarisation and word-level timestamps, then edit any line before rendering
- Apply an optional lip-sync layer billed only on frames where a face is actually talking
- Fan a batch of files out across several target markets in parallel and download the results as a zip
- Drive the whole pipeline programmatically through a REST API with signed webhooks
When to use Vaani / When not to
A quick filter to help you decide if Vaani is the right fit.
When to use Vaani
- YouTubers and independent creators re-voicing an existing back catalogue into Indian languages
- Broadcasters and news publishers localising segments overnight, at broadcast loudness
- EdTech teams scaling a course library across thirteen Indian languages without re-recording anyone
- OTT platforms preparing multi-language masters for regional audiences
- Engineering teams wiring dubbing into a content pipeline through the REST API and webhooks
When not to use Vaani
- Anyone needing a mobile app: Vaani runs in the browser and through its API, nothing else
- Editors expecting a plugin for Premiere, Resolve or any NLE, or a Zapier-style connector
- Teams under a procurement or DPO review, since the site publishes no GDPR statement, no DPA and no hosting location
- Buyers wanting a permanent free tier: the free offer is a one-shot allowance of twenty coins and a single active project
- Footage with no identifiable person speaking on camera, where voice cloning has nothing to work from
How to use Vaani
A typical end-to-end flow, from setup to results.
- Sign in with a Google account in one tap, or ask hello@vaani.media to set up enterprise SSO
- Upload a source video in MP4, up to 1080p
- Let Vaani transcribe the audio with speaker diarisation, so each voice on a multi-speaker edit stays distinct
- Pick a target language and let the scene-aware model translate the transcript
- Review and edit the translation line by line before anything is rendered
- Decide on lip-sync: on for talking-head footage, off for a B-roll-heavy edit
- Check the exact coin cost previewed in the Studio panel before launching the job
- Render, then download the finished MP4 with the cloned voice mixed against the original score
- For volume, switch to Glot: drop several files, fan them out across markets, collect a zip
- For automation, mint an API key in the dashboard, POST a job, then poll or register a webhook
Pros & Cons
Pros
- Cloning works from the footage itself, so there is nothing to record and nothing to train
- Rare depth in Indian languages, including Assamese, Odia and Nepali, next to thirty global ones
- Lip-sync is billed per talking frame, so a thirty-minute video with ninety seconds of talking head costs roughly the ninety seconds
- Every tier runs the same pipeline: the plan changes the price, not the output quality
- The translated transcript is editable line by line before rendering, and the coin cost is previewed before each job
- API and web app draw on one usage ledger, with no separate API meter and no per-seat lock-in
- An explicit commitment not to train models on customer media, plus a card-free trial to test it
Cons
- No identifiable company: no postal address, no legal form, only a footer reading Strible Advantai while the page metadata says Vaani by Advant AI
- Not a single mention of the GDPR, no DPA, no Article 27 representative, no subprocessor list
- No hosting country or region is disclosed anywhere, and no security certification is displayed
- Very young: the domain was registered on 25 April 2026 and first archived on 13 June 2026
- One address, hello@app.vaani.media, carries support, legal and privacy alike
- API access starts at the Studio plan, $299 per month; below it, minting a key returns a 402
- Two pricing units coexist and disagree: the homepage quotes Global from $1.50 per minute while its own rate card lists $2
Pricing & Plans
There is no permanent free plan. The free offer is a one-time trial, requiring no card, worth twenty coins on signup, which covers five minutes of Indian-language dubbing and two minutes of global dubbing on a single active project. The lowest paid entry point is the Creator plan at USD 49 per month, billed monthly and cancellable at any time from the account settings, with the plan remaining active until the end of the period already paid for. Usage beyond a plan's included minutes is charged at the per-minute rate published on the pricing page.
- 20 coins on signup
- 5 minutes Indian and 2 minutes global dubbing
- lip-sync at Creator rates
- Studio and Glot access
- 1 active project
- no card required
- 100 coins per month
- 30 minutes Indian
- 15 minutes global and 5 minutes lip-sync
- dubbing at 2.0 coins per minute
- all languages
- Basic and Pro lip-sync
- unlimited projects
- 1 seat
- 700 coins per month
- 200 minutes Indian
- 100 minutes global and 30 minutes lip-sync
- dubbing at 1.65 coins per minute
- Developer API
- 5 team seats
- priority queue
- 4
- 000 coins per month
- 900 minutes Indian
- 450 minutes global and 150 minutes lip-sync
- dubbing at 1.5 coins per minute
- API and webhooks
- named support
- 15 team seats
- unlimited coins
- negotiated contract rates
- on-premise or private-cloud deployment
- custom voice library
- SLA with named support
- SSO and audit logging
Data, GDPR & hosting
A consolidated view of how Vaani handles your data.
GDPR overview
There is no mention of the GDPR anywhere on the site. The word does not appear once across the privacy policy, the terms or any other page, and there is no compliance claim, no data processing agreement, no Article 27 EU representative, no data protection officer, no subprocessor list and no stated hosting jurisdiction. Some rights are nevertheless offered in practice, simply without the legal framing: the privacy policy of 1 May 2026 lets you export everything held about you, or delete your account and all associated media, by writing to hello@app.vaani.media, with a seven-day response commitment. Material changes to the policy are announced to registered users fourteen days in advance. Treat the silence as a gap to raise with the vendor, not as evidence either way.
Who owns the data?
The terms are unusually direct on ownership. You own every output you generate; Vaani keeps only a non-exclusive licence to host those files for as long as your account stays open, so playback, export and re-download keep working. The company states plainly that it does not train models on your media. The other side of the bargain is that you must own the rights to whatever you upload, or hold written permission from the rightsholder: uploads are not pre-screened, and Vaani says it cooperates with valid takedown requests. The brand, the website, the apps and the pipeline orchestration code remain Vaani's, with no licence to resell or redistribute them.
Reuse rights
Yes. The dubs, transcripts and mixed masters you generate are yours to publish, monetise, edit or redistribute without asking Vaani for anything, commercial use included, and no attribution requirement appears anywhere in the terms. The only licence you grant back is a hosting licence, and it lapses with the account. The real constraint sits upstream rather than downstream: your right to reuse an output is only ever as strong as your right to the source video it came from, since Vaani does not check that for you.
Data retention & training
Hosting summary
Vaani discloses nothing about where data is hosted. No country, region, data centre or cloud provider is named in the privacy policy, the terms or anywhere else on the site, and no jurisdiction clause governs data processing. Source files are described only as living in the account's private storage, with the pipeline reading them, generating stems and language tracks, and writing results back to the same place. Outputs are served from storage.vaani.media and the API from api.vaani.media, neither of which reveals a location. The site itself sits behind Cloudflare on an anycast address, which is a network observation rather than a statement by the vendor and says nothing about where processing or storage happens. No subprocessor list is published, so the third parties involved in the pipeline — the lip-sync engines are named on the homepage, the payment provider is visible in the checkout — are not formally disclosed as data processors. Anyone with a data residency requirement should ask before uploading.
Things to keep in mind
Risks and trade-offs to weigh before adopting Vaani.
- You alone carry the rights question: uploads are not pre-screened, so dubbing footage you do not own puts you, not Vaani, in front of a takedown or a claim
- Cloning a recognisable voice into a language its owner never spoke is an ethical decision, not a technical one; consent from the person on camera is not something the platform asks for
- No hosting jurisdiction is disclosed and the GDPR is never mentioned, so you cannot tell a regulator where your footage was processed
- A single address handles support, legal and privacy, and no phone number or postal address exists as a fallback if it goes unanswered
- Liability is capped at what you paid over the previous twelve months, and no specific dub quality is guaranteed
- Failed runs caused by malformed or unsupported source media are not refunded, so a large batch can burn coins without producing anything usable
- Fluent machine dubbing invites you to skip the native-speaker review; scene-aware translation reduces the risk of a cultural misfire but does not remove it, and errors ship at the speed of the pipeline
Setup & Integrations
Technical difficulty
Very low on the product side. Nothing to install, everything in the browser: sign in with Google, upload an MP4 up to 1080p, pick a language, render. No voice samples to prepare and no training step. Reviewing the translated transcript line by line is available but optional. The API asks more: mint a key whose scope is fixed at creation and cannot be changed, send a Bearer header, POST an asynchronous job, then poll or verify an HMAC-SHA-256 signature on the raw webhook body. Note that API access only unlocks from the Studio plan upwards.
Deployment
Supported languages
Behind Vaani
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What does Vaani actually do?
Do I have to train the voice clone or record samples first?
Which languages are supported?
How much does it cost, and is there a free plan?
How is lip-sync billed?
Is there an API?
Does Vaani train its models on my videos?
Who owns the dubbed output?
How long is my data kept, and how do I delete it?
Is the company behind Vaani identified?
Should you pick Vaani?
Vaani makes one bet and makes it well: the voice you hear in the dub should be the voice already on the tape. Cloning from the source removes the sample-collection step that slows most dubbing workflows, and the depth in Indian languages — thirteen of them, Assamese and Odia included — is genuinely hard to find elsewhere. Two design decisions stand out for anyone costing this at volume: lip-sync is billed per talking frame rather than per minute of video, which makes B-roll-heavy edits far cheaper than a flat rate would suggest, and every tier runs the same pipeline, so paying more buys a lower unit cost rather than a better dub.
The reservations are not about the product. They are about the company behind it. There is no postal address, no legal form, no registration number; the footer says one name and the page metadata says another. The GDPR is never mentioned, no data processing agreement is offered, no subprocessor list is published, and no hosting country is disclosed — on a tool whose whole job is to ingest video that may carry other people's likenesses and voices. The domain was registered in April 2026 and first archived in June, so there is very little track record to lean on either.
For an individual creator, the free trial costs nothing to run and settles the quality question in an afternoon; the Creator plan at $49 is a modest commitment. For a broadcaster, an EdTech team or anyone whose procurement passes through legal, the sensible move is to ask directly for the hosting jurisdiction, a DPA and a company registration before uploading anything sensitive. The technology looks ready; the paperwork is not there yet.
- Choosing a selection results in a full page refresh.
- Opens in a new window.