Synthesia public-data review: what the credits actually buy
Synthesia bills in credits at a flat 2 credits per second — 120 per finished minute — and that rate does not change with the avatar you pick. So the allowances convert cleanly: Starter's 1,200 monthly credits are 10 minutes, Creator's 3,600 are 30. The two numbers that actually move your bill are hidden elsewhere: credits never roll over, and API generation costs 75 credits a minute against 120 in the editor — making Synthesia's API cheaper than its own app, the exact reverse of HeyGen.
The official plans
These are Synthesia's own published figures from its pricing page, checked September 2026.
| Plan | Price | Credits | Stated video | Seats | Avatars |
|---|---|---|---|---|---|
| Basic (free) | $0 | 1,200/mo | ~10 min/mo, watermarked | 1 editor | 9 stock |
| Starter | $29/mo, or $18/mo yearly | 1,200/mo, or 14,500/yr | 10 min/mo, or 120 min/yr | 1 editor + 3 guests | 125+ stock, 3 personal |
| Creator | $89/mo, or $64/mo yearly | 3,600/mo, or 44,000/yr | 30 min/mo, or 360 min/yr | 1 editor + 5 guests | 180+ stock, 5 personal |
| Enterprise | Contact sales | Custom | Unlimited minutes | Custom | 240+ stock, unlimited personal |
The free Basic plan is a demo, not a starter tier: output carries a Synthesia logo and, critically, cannot be downloaded as MP4 at all. Starter is the first plan that removes the logo and allows downloads, and adds AI Dubbing. Creator is the first plan with API access, interactive videos and branded video pages. Enterprise adds SAML/SSO, live collaboration, Brand Kits and one-click translation into 80+ languages.
The flat rate that decides your bill
Synthesia's credits help page puts standard video at 2 credits per second — 120 per minute — and that rate is the same whichever avatar you use. Other features bill at their own rates.
| Generation type | Credits |
|---|---|
| Standard video | 120 per minute (2/sec) |
| AI Dubbing — with lip sync | 50 per minute |
| AI Dubbing — without lip sync | 25 per minute |
| Bulk Personalization | 90 per minute |
| API usage | 75 per minute |
| Veo 3.1 video asset — with audio | 192 |
| Veo 3.1 video asset — without audio | 96 |
| Veo 3.1 Fast — with audio | 60 |
| Veo 3.1 Fast — without audio | 48 |
| Gemini Omni video asset | 48 |
| FLUX.2 AI image | 3 |
| Nanobanana 2 / OpenAI GPT Image 2 | 9 each |
The third-party video models are priced per generated asset rather than per minute, and the help page does not state their clip length, so cost per second cannot be derived for those rows. If you want Google's own per-second rates for the same engine, the Veo profile has them.
What that means per plan
Dividing each allowance by each rate gives the finished output. The arithmetic below is CartSignal's, applied to Synthesia's published numbers.
| Workload | Starter monthly (1,200) | Creator monthly (3,600) | Starter yearly (14,500) | Creator yearly (44,000) |
|---|---|---|---|---|
| Standard video (120/min) | 10 min | 30 min | ~121 min | ~367 min |
| Dubbing, lip sync (50/min) | 24 min | 72 min | 290 min | 880 min |
| Dubbing, no lip sync (25/min) | 48 min | 144 min | 580 min | 1,760 min |
| Bulk Personalization (90/min) | ~13 min | 40 min | ~161 min | ~489 min |
| API (75/min) | — | 48 min | — | ~587 min |
The dubbing rows are the ones worth reading twice, because they run the opposite way to most people's expectation. Dubbing costs less than generating — 50 credits a minute with lip sync against 120 for standard video, and 25 without. So a Starter plan bought to localise an existing video library delivers about 24 lip-synced minutes a month, not ten, and turning lip sync off doubles that again. If you already have a presenter on camera, dubbing is much the cheaper path.
Corrected 11 September 2026: an earlier version of this page listed these rates as 240 and 120 credits per minute and described dubbing as costing double standard video. That was a misreading of the help page — 240 is the Basic plan's monthly dubbing allowance, not a per-minute rate. The rates above are the figures Synthesia's credits guide publishes, and the conclusion reverses.
Three things the plan table hides
Credits do not roll over — at all. Synthesia's help page is blunt: credits "are reset at the start of the billing period and there is no accumulation of unused credit." A quiet month is simply forfeited. That is stricter than HeyGen, which carries unused credits one extra cycle, and stricter than Murf, which does the same with Voice Generation Time. If your output is lumpy, this rule matters more than the sticker price.
Creator costs more per credit than Starter. Price divided by allowance gives about $0.0242 per credit on monthly Starter, $0.0247 on monthly Creator, $0.0149 on yearly Starter and $0.0175 on yearly Creator. So Creator is about 2% more expensive per credit billed monthly and about 17% more expensive billed yearly. Almost every SaaS ladder gets cheaper per unit as you climb; this one gets dearer. Creator buys API access, interactive video, more personal avatars and priority support — capabilities, not cheaper volume. If you only need more minutes, two things are worth pricing before you upgrade.
The annual discount is a discount, not a bigger allowance. Yearly Starter's 14,500 credits are only 100 more than twelve monthly allocations of 1,200; yearly Creator's 44,000 beats 12 × 3,600 by 800. The saving is in the price — about 38% per credit on Starter and 29% on Creator — not the volume. The secondary benefit is pooling: the pricing page quotes annual plans as minutes "per year" rather than per month, so a year's work is not chopped into twelve expiring buckets. Because the reset rule is written per billing period, confirm with Synthesia whether your annual pool is released once yearly or in monthly slices, since with no rollover that distinction decides whether a heavy month is even possible.
In cost-per-finished-minute terms, standard video runs about $2.90 on monthly Starter, $2.97 on monthly Creator, $1.79 on yearly Starter and $2.09 on yearly Creator.
The API is cheaper than the app
API generation is rated at 75 credits per minute against 120 in the editor — 37.5% less, or 1.6x more finished minutes from the same pool. There is no separate API bill; it draws on the subscription credits you already have, and access begins on Creator.
This is worth flagging because it inverts the pattern elsewhere in this directory. HeyGen's API is a standalone pay-as-you-go meter that no subscription includes, and it runs several times more per minute than the equivalent credits. ElevenLabs likewise splits its API onto a separate meter. Synthesia charges one pool and discounts the programmatic route. For a team already generating videos from a script pipeline, that flips the usual build-versus-click calculation: at Creator's yearly rate an API minute costs about $1.31 against $2.09 in the editor.
Express-2 is included, not an upsell
Synthesia announced Express-2 on 4 September 2025 as a new video and voice engine: a diffusion transformer model driving full-body avatars that gesture, paired with an in-context voice cloning system that preserves accent and rhythm. It outputs 1080p at 30fps with no stated limit on video length.
The commercially interesting part is the billing. Synthesia's release notes state Express-2 avatars are "live across all plans—with no limits and no extra cost," so they consume the same flat 120 credits per minute as anything else. Compare that with HeyGen, where the flagship Avatar V burns 48 credits per minute against Avatar III's 4 — a 12x premium for the newest model, plus a 3-minute cap per video. Synthesia's newest model carries neither a premium nor a length cap. What does vary by plan is the roster: free Basic users get Zola and Michael, while paid plans add Ryan, Ada and Ellie.
Synthesia's quality claim for the voice engine — that in blind tests with 100 native English-speaking evaluators across 17 accents, Express-Voice rated highest for matching the original speaker's identity, rhythm and accent — is the company's own figure, not an independent benchmark, and CartSignal has not verified it.
Synthesia 3.0, announced 1 October 2025, added Video Agents that hold real-time two-way conversations, prompt-based avatar environments, personal avatars from a single image, and frame-accurate dubbing into 30+ languages. Copilot and Courses were flagged as coming in 2026; treat unreleased features as unconfirmed until they ship.
Editing an existing video is mostly free
A common worry with per-minute billing is that every revision costs another full render. Synthesia's video-edit consumption page says otherwise: you are charged only for what changed, not for regenerating the whole video.
- Free: deleting words, reordering existing words, punctuation changes, avatar framing, removing an avatar, changing appearance or outfit, pronunciation fixes, tone changes.
- Charged at 2 credits per second: newly added words, replaced words (the new duration), a newly added or swapped avatar (for the duration it is visible), and a changed voice language or accent.
Synthesia's own example: a 2:40 video costs 320 credits, and adding one second of new script costs 2 more, for 322. That makes iteration genuinely cheap and is a real advantage for training content that gets revised as policies change — the use case Synthesia is built around. Note the one asymmetry: switching a voice's language or accent is charged across the affected duration, so multilingual experiments are not free the way cosmetic edits are.
What Synthesia actually does
- Avatar-led video from a script, with 240+ stock avatars at Enterprise and personal avatars creatable from a single image.
- Express-2 full-body avatars with gesturing and expression, on every plan at no premium.
- Voice cloning via Express-Voice, across 160+ languages and voices.
- AI Dubbing with or without lip sync, at 50 or 25 credits per minute — cheaper than generating video.
- Interactive video — clickable elements, branching, and Video Agents that listen and respond — from Creator up.
- Third-party generative models in-platform: Veo 3.1, Gemini Omni, FLUX.2, Nanobanana 2 and GPT Image 2, billed per asset from the same pool.
- API access from Creator, at a 37.5% discount to editor rates.
Best-fit use cases
- Corporate training and onboarding libraries that get revised often, where the cheap-edit rule compounds
- Internal communications and policy updates needing a consistent presenter without filming
- Product walkthroughs and explainers where a fixed presenter is fine and length is unpredictable — there is no per-video cap
- Script-pipeline video generation at volume, where the API's 75 credits per minute is the cheapest route on the price list
- Localising an existing video library, where dubbing at 50 credits a minute costs less than half the rate of generating new footage
Limitations to verify
- Credits never roll over; unused allowance is forfeited at each billing reset
- Creator costs more per credit than Starter — upgrade for features, not volume
- The free Basic plan cannot download MP4s and watermarks output
- Lip sync doubles the dubbing rate (50 vs 25 credits a minute), though both stay below the 120 charged for standard video
- API access starts on Creator; Starter has none
- Third-party model assets (Veo 3.1, Gemini Omni) are priced per asset with no published clip length
- Whether an annual plan releases credits yearly or monthly should be confirmed before committing
- Enterprise pricing is unpublished, so the jump from Creator cannot be costed from public data
- Avatar likeness and voice-consent policy should be read directly before cloning anyone but yourself
How it fits against other tools here
The closest head-to-head is HeyGen, and the pricing structures differ more than the feature lists do. HeyGen spreads cost across avatar models — 4 to 48 credits per minute — so the model you pick swings the bill twelvefold, and its API is a separate, pricier meter. Synthesia charges one flat rate for any avatar and discounts its API, but never rolls credits over. Broadly: HeyGen rewards picking a cheap model, Synthesia rewards predictable volume. The HeyGen vs Synthesia comparison works through that decision, and HeyGen alternatives covers the wider field.
If you want generated footage rather than a presenter, this is the wrong category — Veo, Kling and Runway are the text-to-video engines, and Sora alternatives is the migration guide after OpenAI's September 2026 API removal. Fliki is the budget lane for text-to-video with narration, and Descript is the editor you reach for afterwards. On voice alone, ElevenLabs remains the realism benchmark. The roundups are best AI avatar video tools and best AI video tools. If you publish in the EU, Article 50 labeling duties apply to synthetic avatar video whichever tool made it.
FAQ
How much does Synthesia cost?
Starter is $29/month or $18/month billed yearly; Creator is $89/month or $64/month billed yearly. A free Basic plan gives 1,200 credits a month but watermarks output and does not allow MP4 downloads. Enterprise is custom-priced with unlimited video minutes.
How many credits does a minute of Synthesia video cost?
120 — two credits per second, the same for any avatar. So Starter's 1,200 monthly credits are 10 minutes and Creator's 3,600 are 30. Other features differ: AI Dubbing with lip sync is 50 credits per minute, without lip sync 25, Bulk Personalization 90, and API generation 75. Dubbing existing footage is therefore cheaper than generating new video.
Do Synthesia credits roll over?
No. Synthesia's help page states credits reset at the start of the billing period with no accumulation of unused credit, so monthly billing is use-it-or-lose-it. Annual plans are quoted as a yearly allowance instead — 14,500 credits on Starter, 44,000 on Creator.
Is the Synthesia API cheaper than making videos in the app?
Yes, by 37.5% per minute — 75 credits against 120 — so the same pool yields 1.6x more minutes. There is no separate API bill; it draws on your subscription credits, and access starts on Creator. That is the reverse of HeyGen, whose API is a standalone and more expensive meter.
Does Synthesia charge extra for Express-2 avatars?
No. Synthesia's release notes state Express-2 is live across all plans with no limits and no extra cost, so it bills at the same flat 120 credits per minute. The roster varies by plan: Basic gets Zola and Michael, paid plans add Ryan, Ada and Ellie.
Has CartSignal tested Synthesia?
No. This is a public-data review built from Synthesia's official pricing page, credits help centre articles, video-edit consumption article and the Express-2 and Synthesia 3.0 announcements, checked in September 2026. No hands-on render quality, lip-sync or likeness testing is claimed, and vendor benchmark figures are labelled as vendor claims.
Source links
- Official Synthesia pricing page (plan prices, credit allowances, seats, avatar counts)
- Official credits guide for self-serve plans (consumption rates, no-rollover rule)
- Official credit consumption for video edits (what re-rendering charges)
- Official Express-2 announcement, 4 September 2025 (vendor benchmark claims)
- Official Synthesia 3.0 announcement, 1 October 2025
- Official Synthesia homepage