Per-minute list prices are the easiest numbers in voice AI to compare and the least useful. Nobody pays list price: you pay a monthly plan plus whatever your minutes spill over it, and the plan decides the real bill. So here is the real bill, at five monthly volumes, for two platforms whose pricing pages publish enough to do the math: ElevenLabs Agents and Lokutor.
This comparison is deliberate. Lokutor is our product, and the math favors it at every volume below. The numbers are not ours: every ElevenLabs figure below is from elevenlabs.io/pricing, read 27 September 2026. Do your own math before you believe anyone's, including this post.
The two plan tables
ElevenLabs Agents, monthly billing, from their pricing page on 27 September 2026:
| Plan | $/month | Agent minutes included | Effective per minute | Additional minute |
|---|---|---|---|---|
| Free | $0 | 15 | - | $0.080 |
| Starter | $6 | 75 | 8.0c | $0.080 |
| Creator | $22 | 275 | 8.0c | $0.080 |
| Pro | $99 | 1,238 | 8.0c | $0.080 |
| Scale | $299 | 3,738 | 8.0c | $0.080 |
| Business | $990 | 12,375 | 8.0c | $0.080 |
One thing worth crediting: 8 cents a minute, flat, on every tier. No volume discount, but no games either. Burst pricing when you exceed your concurrent call limit is $0.160 a minute, and the LLM (Gemini 2.5 Flash in their estimator) bills separately from credits. Telephony is excluded.
Lokutor, from the pricing post published 23 September 2026:
| Plan | $/month | Minutes included | Effective per minute | Overage |
|---|---|---|---|---|
| Free | $0 | 60 | - | - |
| Starter | $19 | 1,000 | 1.9c | 2c |
| Growth | $99 | 6,000 | 1.65c | 1.8c |
| Business | $349 | 25,000 | 1.4c | 1.5c |
The Lokutor minute includes speech recognition, the language model, and the voice. Phone minutes count double inbound and quadruple outbound, because carriers charge more to call a mobile.
The monthly bill at real volumes
Cheapest legal combination of plan plus overage on each side, voice agent minutes only:
| Minutes per month | ElevenLabs Agents | Lokutor | Ratio |
|---|---|---|---|
| 500 | $40 (Creator + 225 over) | $19 (Starter) | 2.1x |
| 2,000 | $160 (Pro + 762 over) | $99 (Growth) | 1.6x |
| 10,000 | $800 (Scale + 6,262 over) | $349 (Business) | 2.3x |
| 50,000 | $4,000 (Business + 37,625 over) | $724 (Business + 25,000 over) | 5.5x |
At 100,000 minutes the ElevenLabs bill is $8,000 (Business plus 87,625 overage minutes). Lokutor's published overage covers up to one extra month of plan minutes, so 100,000 is volume-contract territory; the first 50,000 of it costs $724.
The gap widens with volume because the overage rates do all the work: 8 cents against 1.5.
What this math leaves out, honestly
Concurrency. ElevenLabs Business allows 40 concurrent calls, Lokutor Business 20. If your peak is 30 simultaneous calls, the cheaper minute does not help you.
Languages. ElevenLabs lists 32. Lokutor lists 9: Spanish, Catalan, Galician, Basque, Portuguese, English, French, Italian, German. The Iberian set is first-class, which is the point of the product, but 9 is fewer than 32.
Catalog and features. ElevenLabs offers thousands of library voices, dubbing, sound effects, and a music product. Lokutor offers 10 voices and voice cloning on upper plans. If your product depends on a specific ElevenLabs voice, a per-minute table is not the decision.
The LLM line. On ElevenLabs the language model bills from credits on top of the minute. Their estimator puts Gemini 2.5 Flash at $0.34 for 270 minutes, so it is small, but it is not zero. On Lokutor it is inside the minute.
Free tiers. ElevenLabs gives 15 agent minutes a month. Lokutor gives 60, no card. For evaluation, both are enough to hear your own use case, which is the only benchmark that matters.
Where the difference comes from
Lokutor's voice model is a few tens of millions of parameters and runs on CPUs; the category norm is models tens of times larger on GPUs, priced to match. The 1.4 to 2 cent minute is what the architecture costs to run, plus a business. Whether that trade is worth it depends on the section above, not on the headline number.
The latency side of the same architecture is in Voice agents on CPU vs the GPU incumbents, and the migration mechanics are in Migrate from ElevenLabs to a CPU voice API in about 20 lines. Pricing sources: ElevenLabs pricing, Lokutor pricing, both as of late September 2026. Try the free tier at app.lokutor.com.
Try the voices in your browser on Hugging Face. Lokutor 2.0 launches on Product Hunt October 13.
Top comments (0)