🎓 Synthesia AI Review 2026: Credits, Avatars & Verdict

Every roundup currently ranking for Synthesia's pricing tells you the same thing: Starter buys ten flat minutes a month, Creator buys thirty. Synthesia's own pricing page stopped working that way. Self-serve plans now run on a shared credit pool, and the conversion rate is different for every feature that draws on it. If you budgeted in minutes, you budgeted wrong.

At a glance#

Credits replaced minutes, and the rate is not uniform#

The unit is the credit. Synthesia's knowledge base puts standard video at 2 credits per second, so a finished minute costs 120 credits. That part is simple. The rest is not, because the same balance pays for dubbing, API renders, bulk personalisation and generated images, each at its own rate.

Read that as a budget rather than a spec sheet. A Starter plan holding 1,200 credits a month is ten minutes of video, or five minutes of lip-synced dubbing, or some mixture of the two. It is not ten minutes plus dubbing on top. Every review still describing this as "10 video minutes a month" is quoting a number that only holds if you never touch the localisation feature Synthesia sells the whole platform on.

The generated-asset rates matter for the same reason. An image from FLUX.2 costs 3 credits, and one from Nano Banana 2 or GPT Image 2 costs 9. Custom video assets built with Gemini Omni or Veo 3.1 run between 48 and 192 credits each depending on the spec. Those are small numbers individually and a meaningful hole in a 1,200-credit month collectively.

So what: price the entire job, translation passes and B-roll included, before you pick a tier.

The API costs less than the editor for the same minute#

Here is the part nobody has drawn out of the rate card. A minute rendered through the API draws 75 credits. The same minute rendered in the web editor draws 120. That is a 37.5% discount for identical output, sitting inside the Creator plan, which is exactly where API access begins.

Run it forward. Creator's annual allocation is 44,000 credits. Through the editor that is 366 finished minutes. Through the API it is 586. Same subscription, same money, 220 extra minutes of video decided entirely by where you pressed render.

This only helps teams who can drive a pipeline from a script file or a spreadsheet, which is most L&D teams producing the same module in eight languages and almost no marketing teams shipping one hero video a quarter. If your work is repetitive and templated, the editor is the expensive habit.

So what: on Creator with any real volume, moving generation to the API is worth roughly 60% more output for zero extra spend.

What a finished minute actually costs#

Monthly billing is the expensive mistake here, and it costs more than the sticker gap suggests.

Starter at $29 a month for 1,200 credits works out to $2.90 per finished minute. Pay annually at $18 a month and you get 14,500 credits for the year, which is 120 minutes for $216, or $1.79 a minute. Identical plan, identical features, 38% cheaper per minute for committing twelve months up front.

Creator is flatter. At $89 monthly you get 3,600 credits, 30 minutes, $2.97 a minute. Annual at $64 a month buys 44,000 credits for $768, giving 366 minutes at $2.09. Route those through the API instead and the same $768 covers 586 minutes, or $1.31 a minute, which is the cheapest honest rate anywhere on the platform.

Then the reset rule takes some of it back. Credits do not roll over: Synthesia's documentation says the balance resets at the start of each billing period with no accumulation. A quiet month is money burned, so the plan should be sized against your slowest month, not your best one. That single line is why the annual maths above is a ceiling rather than a forecast.

So what: annual billing, sized to your floor, with the API wherever you can reach it.

Re-rendering is cheaper than the reviews claim#

One correction runs the other way, in the buyer's favour. The highest-ranking pricing article on this topic warns that re-rendering after edits counts against your allotment. Synthesia's own documentation contradicts that.

Charges trigger only on edits that add something: new words in the script, a replaced word, adding an avatar where there was none, swapping one avatar for another, changing voice language or accent. Edits that add nothing are free. Deleting words, reordering them, fixing punctuation, switching an avatar from full body to a circle crop, dropping the avatar for a voice-only cut, changing clothing, adjusting pronunciation, retuning voice tone. None of those recharge.

That flips the standard workflow advice. The cautious approach of writing short and adding later is the one that costs you. Write long, generate once, then cut down inside the editor, because trimming is free and rewriting is not.

So what: over-write the first draft deliberately, then subtract.

Express-2 and what else shipped in 3.0#

Express-2 is the engine behind the avatars, pairing voice cloning with full-body gesture generation so a presenter points, gestures and shifts weight instead of sitting rigid from the shoulders up. Micro-expressions, meaning eyebrow raises and breathing tied to script emotion, are the detail that holds up past the third minute of a training module where a static face starts to get noticed. It runs across paid plans rather than being gated to Enterprise.

Three other pieces from the 3.0 release get skipped in most reviews. Video Agents are scripted avatars that respond inside an interactive branch rather than playing linearly. Courses mode bundles several videos into one graded module. Copilot rewrites a script to hit a target run time instead of making you trim by hand. None is a headline feature. Together they are the difference between an avatar video tool and a training-program tool that happens to use avatars, which is the category Synthesia is actually competing in.

AI Playground is the one worth budgeting for. It puts Veo 3.1 and Sora 2 B-roll generation inside the app, so a training team needing an establishing shot does not pay those models' own API rates separately. It draws from the same credit pool at 48 to 192 credits per asset, so treat it as a convenience with a price rather than a freebie.

By the numbers (checked August 2026)#

Figures come from synthesia.io's August 2026 pricing page and Synthesia's own knowledge-base articles on credits and on credit consumption for video edits. Confirm before you buy: these tiers have moved more than once this year, and the shift from minutes to credits is exactly the kind of change that leaves published reviews wrong.

Who should pay for this#

L&D teams, compliance and onboarding programmes, and any organisation shipping the same module in a dozen languages without hiring a dozen translators. The enterprise case is not avatar realism, it is SSO/SAML, SCORM export for LMS integration, and a credit budget a finance team can forecast a year out. Synthesia reports adoption by more than 90% of the Fortune 100, which is a claim about procurement rather than about pixels.

Skip it if you are a solo creator watching cost per video. HeyGen's credit system stretches further at low volume even with its own traps, and our Synthesia vs HeyGen test breaks down where each one wins. If dubbing is the real job rather than avatar presenting, the wider field in our AI dubbing and translation tools guide is worth reading before you commit a year of credits. And if you have not yet decided whether an avatar platform is the right category at all, start with how to create an AI avatar, then check the per-minute numbers against the rest of the market in our AI video cost breakdown.

Our rating: 4.4/5. Express-2 and the 3.0 stack make this the more complete platform for anyone billing a training budget, and the credit rate card is published in enough detail that you can model a year of spend before you sign, which is rarer in this category than it should be. It loses half a point for a pool that expires monthly and charges double for the localisation feature its own marketing leads with.

Official resources#

• Synthesia official site: free plan with 1,200 credits a month, no card required
• Synthesia pricing: current tiers and credit allocations
• Synthesia Academy: official tutorials and template walkthroughs

Go deeper#

• Synthesia vs HeyGen: the full head-to-head with a per-use-case verdict
• How to create an AI avatar: the workflow before you pick a platform
• AI dubbing and translation tools: the localisation field, mapped
• AI video cost breakdown: real per-minute maths across the market
• Colossyan vs Synthesia: the L&D match-up, and why Colossyan's widely-quoted $27 plan no longer exists
• Best AI UGC and avatar video tools: where Synthesia sits against Arcads, Creatify, HeyGen and Captions

Prefer video? Hand-picked walkthroughs#

Reading is faster, but if you want to see it done, these are the best tutorials we vetted for this topic: