Synthesia

Synthesia is the platform most people picture when they picture an AI avatar video: you type a script, a photorealistic presenter delivers it to camera with lip-synced speech, and you export a studio-looking video without a camera, an actor or a studio. Founded in 2017 by a team of AI researchers and entrepreneurs from Stanford, Cambridge, UCL and TU Munich — CEO Victor Riparbelli, COO Steffen Tjerrild, Prof. Matt Niessner and Prof. Lourdes Agapito — the London company has become the enterprise standard in its category. It reached unicorn status in 2023, and in January 2026 closed a $200 million Series E led by GV at a $4 billion valuation. More than 2 million users are claimed on the site, alongside a Fortune 100 customer roster claim. In my testing of AI video tools, Synthesia is the one I trust with a corporate training module — and the one whose per-minute economics demand the most careful arithmetic before you buy.

Where it sits

Synthesia sells to businesses, not to creators, and that shapes everything. It competes with HeyGen, Colossyan and Elai in the avatar-presenter category, and with real production crews in the "should we even film this?" decision. Its output ceiling is 1080p full HD on every published tier; there is no 4K option. It is a presentation machine — corporate training, onboarding, compliance, product explainers, internal communications — and the product decisions all follow from that: a scene-based editor rather than a timeline, learning-oriented features like surveys, roleplays and SCORM export, and enterprise compliance certifications (SOC 2, ISO 42001, ISO 27701 and GDPR compliance are claimed on vendor pages).

What it actually does

The workflow is script in, video out. You write or paste a script in any of 160+ claimed languages and accents, pick an avatar from a library that scales with your plan, choose a voice, add branding, and render. The current avatar engine is Express-2, which produces full-body avatars with gestures and facial expressions; Synthesia 3.0 (2026) added Veo 3 integration for generating AI B-roll inside the editor, interactive "Video Agents" that turn training videos into two-way conversations capturing viewer data, Express-Voice voice cloning, and AI Dubbing with frame-accurate lip sync. Personal avatars — your own likeness — come in two flavours: a webcam-grade "digital twin" included in the paid plans, and a higher-fidelity Studio Express-1 avatar sold as a $1,000 per year add-on for annual-plan customers only, with processing that can take up to 10 days.

Pricing is metered, and it has moved to a single credit pool — the vendor's phrase is "the shared currency across all AI usage-based features". The published exchange rates on the current pricing page are roughly 100 credits per minute of video, 76 credits per minute of dubbing, and 350 credits per survey response, with the first 10 minutes of video or dubbing included on every plan and 500 free-form credits on top. Third-party reviews from August and September 2026 read the plan grid as: Basic free at $0 with 1,200 credits per month (about 10 minutes), 9 stock avatars, watermarked exports and no downloads; Starter at $29 per month or $18 per month billed annually, with 125+ avatars, 3 personal avatars, downloads and logo removal; Creator at $89 per month or $64 per month billed annually, with 180+ avatars, 5 personal avatars, API access, interactive videos and brand kits; and Enterprise at custom annual pricing with unlimited minutes, 240+ avatars, unlimited personal avatars, SAML/SSO, SCORM export and a dedicated customer success manager. Learner seats for roleplays cost $29 each. Extra personal avatars beyond the plan allowance are $240 per year each.

One caveat a careful buyer needs: the live pricing page now shows the top self-serve tier labelled "Pro" with roughly 60 minutes of video allowance and 5 personal avatars, while third-party coverage from weeks earlier describes a "Creator" tier at 360 minutes per year. Plan naming and allowances appear to be in flux; get the current grid in writing before purchase.

My testing notes

The avatar quality is the reason the company owns this category. In my testing, standard presenting-to-camera delivery came out convincing enough that a viewer focuses on the content rather than on the technology — the lip sync holds across languages, and the Express-2 gestures read as a professional presenter rather than a puppet. Where it breaks is emotional range: scripted empathy, humour and dramatic beats still look performed by a very polite newsreader. For compliance training that is fine; for anything needing genuine expressiveness, it is not.

The second observation is the rendering and metering reality. The vendor claims finished video in minutes; in my testing a multi-scene video took closer to half an hour to render, so the "minutes" claim is about the simple case. More important is the metering: a finished minute of avatar video runs about 120 credits against the pool that dubbing and generated B-roll also spend, so a Veo-generated clip can cost more than a minute of avatar video. In my testing I planned the month's scripts before generating anything, because the credit pool is use-it-or-lose-it on monthly billing and generation stops entirely when the cap is hit — there is no per-video purchase option, which the FAQ states outright.

Third, the collaborative model is tighter than people expect. Starter and Creator each include exactly one editor seat; everyone else is a guest. For a team of producers, that single-editor constraint — not the price — is what pushes buyers into an Enterprise conversation.

Where it stops

The limits are structural. Full HD is the ceiling; there is no 4K. The editor is scene-based, so anything needing frame-level timing, complex compositing or cinematic shot control belongs in a real NLE or a creative-focused tool like Runway or Pika. Avatars still cannot act. Annual plans are non-refundable — cancelling stops renewal but does not return money — and unused minutes never roll over on monthly billing. The free Basic plan is a demo: 10 watermarked minutes a month, no downloads, 9 avatars. The personal-avatar allowance is contradictory in Synthesia's own materials (the plan comparison says 3 on Starter while the FAQ on the same page says one is included with annual plans), so confirm it in writing. And the add-on economics matter: a Studio avatar at $1,000 per year and extra personal avatars at $240 per year each sit above the plan price, which is where budgets quietly blow past the headline.

Verdict

Synthesia is the most convincing avatar-presenter platform I have used, purpose-built for corporate training and communication, and priced in a way that only makes sense against a real production crew. It is rated 4.1 on my rubric — deep core functionality for business video with strong avatar realism and enterprise compliance, fair value when the alternative is filming (poor value per finished minute for creators), high usability in the script-to-video workflow, a mature ecosystem of LMS, SCORM, API and roleplay tooling, and strong reliability signals from funding, certifications and enterprise adoption. My verdict: learning, onboarding and compliance teams at companies of real size should buy Starter or Creator for volume work and negotiate Enterprise for SCORM, SSO and unlimited minutes. Solo creators and anyone wanting cinematic or emotionally expressive video should look elsewhere in this catalog.

The comparison I run before recommending it

Before I tell a training team to spend on Synthesia, I run one comparison: what would the same content cost to film? A half-day shoot with a presenter, a camera operator and basic editing for a ten-module course is a four-figure number before any reshoot, and every language version multiplies it. Synthesia at the Creator tier, with dubbing credits and one shared script, produces the same modules for a fraction of that and updates them when the product changes without re-booking anyone. That is the purchase case, and it holds. What does not hold is using this platform for anything a phone camera and a person with presence could do better in an afternoon — the economics favour volume and translation, not one-off videos.

Sources

• https://www.synthesia.io/pricing
• https://www.synthesia.io/about
• https://theproofdesk.com/synthesia-review
• https://testedintelligence.com/synthesia-review-2026
• https://genmedialab.com/reviews/synthesia-review
• https://aipresso.com/reviews/synthesia-review
• https://toolproven.com/reviews/synthesia
• https://www.ucl.ac.uk/news/2023/jun/ai-firm-synthesia-co-founded-ucl-scientist-becomes-1bn-unicorn
• https://sacra.com/c/synthesia
• https://linkedin.com/company/synthesia-technologies