Tavus Starter vs Growth: break-even is about 1,014 minutes a month
Tavus Starter is $59 plus $0.37 a minute over 100; Growth is $397 with 1,250 minutes. The crossover, and when a rendered Sume clip fits better than minutes.

On Tavus's published developer plans, Growth becomes cheaper than Starter once you use about 1,014 conversation minutes in a month: Starter is $59 plus $0.37 for each minute past its 100 included, Growth is $397 with 1,250 included. Below that, Starter wins on price. This is arithmetic on the vendor's own page, not a quote, and it covers live conversational video, which is not the same product as a rendered clip.
Figures come from Tavus pricing, read 2026-10-06. Tavus's tier names and numbers can change, so treat the table as a dated snapshot and check the page before you sign anything. Tavus also lists a Basic plan at $0 with 25 minutes and an Enterprise plan with custom pricing, which this comparison leaves out.
How is the break-even calculated?
Starter cost for m minutes, when m is above 100, is 59 + 0.37 x (m - 100). Growth is a flat 397 until 1,250 minutes, then 0.32 a minute past that. Setting Starter equal to 397 gives 0.37 x (m - 100) = 338, so m - 100 = 913.5 and m is about 1,013.5. Rounded up to whole minutes, that is 1,014.
The calculation assumes you use every minute you pay for and that overage is billed per minute as listed. It ignores concurrency, which the page lists as 3 streams on Starter and 10 on Growth, and replicas, which it lists as 3 custom on Starter and 7 on Growth. Either of those can force the upgrade before the money does.
| Minutes in a month | Starter total | Growth total | Cheaper |
|---|---|---|---|
| 100 | $59.00 | $397.00 | Starter |
| 500 | $207.00 | $397.00 | Starter |
| 1,000 | $392.00 | $397.00 | Starter |
| 1,014 | $397.18 | $397.00 | Growth, by 18 cents |
| 1,250 | $484.50 | $397.00 | Growth |
What changes if you render clips instead?
Conversation minutes bill for a live session: someone is talking to the avatar. Many teams buy those minutes to deliver prepared messages, such as a welcome, a product explainer, or a reply to a support ticket. For that, a rendered clip is billed once per job, by seconds, and then reused for every viewer.
On Sume, an Avatar 1.0 talking video takes a script and an avatar handle and must estimate to 4 to 60 seconds (Generate avatar video, read 2026-10-06). Cost depends on the quality tier, so use the live catalog and the cost-by-tier post rather than a number in an article. The relevant difference is shape: a viewer count that grows does not grow a rendered clip's bill, while it does grow live minutes.
Which should you pick?
Do the arithmetic on your own forecast before trusting any crossover. Take your expected minutes for the next three months, add a margin for launch spikes, and price both plans on the busiest month rather than the average one, because overage is billed in the month it happens. If one month alone pushes you past 1,014, the annual picture can still favor Starter.
A short worked case: a team that sells a 20-second welcome message to 5,000 new users a month does not need 5,000 conversations. It needs one clip, maybe a few language variants, and a way to attach it. The live-versus-job comparison walks that split in more detail, and per-minute pricing for avatar video shows how to turn seconds into a monthly figure.
- Pick live minutes when the avatar must hear and answer a person, and budget with the crossover above.
- Pick rendered clips when the message is known ahead of time, when you want captions or a trim afterwards, or when many people watch the same message.
- Pick neither plan size by price alone: concurrency of 3 versus 10 streams decides whether a campaign launch queues up.
- On Sume, concurrency is plan-based and extra jobs wait as queued rather than failing, per Generation admission.
Sources
Related posts
More in Comparisons
- Cost per minute of AI narration: MAI-Voice vs Sume at 900 characters
A minute of narration is about 900 characters. That is $0.0198 on MAI-Voice-2.1, $0.0135 on Flash and $0.0428 on Sume TTS, before the 5.5% fee. Worked table.
- Titan color-guided generation: 1 to 10 hex codes vs a Sume prompt
Titan Image Generator v2 takes 1 to 10 hex colours with a prompt. Sume has no palette field, so colour goes in the prompt. Write it and check it in Python.
- Titan Image Generator v2: 1,408 px and 5 MB limits vs Sume edits
Titan Image Generator v2 caps edit inputs at 1,408 by 1,408 px and 5 MB. How to resize before Bedrock, and what a Sume edit call needs instead.
- Titan vs Nova Canvas image masks: which colour is edited
Titan says mask value 0 (black) is regenerated and bans alpha. Nova Canvas inpainting edits black too. Build one mask check and test any mask on Sume.
Written by Sume