How Voice AI Pricing Actually Works: A Guide to Credits and Tiers
Why do voice AI APIs charge per character? What does a credit actually cost? And how do you pick the right tier for your use case? A practical breakdown of how credit-based TTS pricing works — and how to get the best value.
Why Credits Instead of Just Charging Per Request?
When developers first encounter a credit-based voice API, the question is always the same: why not just charge me per request?
The answer is that "per request" is a terrible unit for TTS billing. A single API call might synthesise a two-word button label or a 40-minute audiobook. The cost of synthesis is overwhelmingly driven by how much text you convert, not how many times you call the API.
Credits solve this cleanly. One credit represents a fixed unit of compute work. You spend credits based on how many characters you synthesise — nothing more, nothing less.
The 9jaLingo Credit Model
At 9jaLingo, the conversion is simple and fixed:
> 1 credit = $0.001 USD ($1 = 1,000 credits)
Every TTS request is billed at a per-character rate, charged before synthesis begins. The formula is:
credits_charged = character_count × rate_per_charThis is applied across three service tiers:
| Service | Rate per character | Rate per 1K chars | USD per 1M chars |
|---|---|---|---|
| Standard TTS (prebuilt voices) | 0.05 cr | 50 cr = $0.05 | $50.00 |
| TTS + Instant Clone | 0.075 cr | 75 cr = $0.075 | $75.00 |
| TTS + Pro Clone (HD) | 0.10 cr | 100 cr = $0.10 | $100.00 |
For developers more comfortable with token-based pricing (familiar from LLM APIs), the equivalent token rates are:
- Standard TTS: $0.20 per 1K tokens
- Instant Clone: $0.30 per 1K tokens
- Pro Clone: $0.40 per 1K tokens
(1 NLP token ≈ 4 characters, BPE-standard.)
Real-World Cost Examples
Here's what common use cases actually cost at Standard TTS rates:
| Use Case | Approx. Characters | Credits Used | Cost |
|---|---|---|---|
| Short social post | 200 | 10 cr | $0.010 |
| Product announcement | 500 | 25 cr | $0.025 |
| 2-min news bulletin | ~1,800 | 90 cr | $0.090 |
| 5-min podcast segment | ~4,500 | 225 cr | $0.225 |
| 10-min training module | ~9,000 | 450 cr | $0.450 |
| 30-min narration | ~27,000 | 1,350 cr | $1.35 |
| Full audiobook (10 hrs) | ~540,000 | 27,000 cr | $27.00 |
The key insight: most everyday developer tasks are extremely cheap. A short announcement costs less than a cent. A full podcast episode costs around 20–25 cents. Costs only become significant at audiobook or IVR scale — and even there, the per-unit economics are competitive.
Choosing the Right Subscription Tier
Credits are also packaged into monthly subscriptions:
| Plan | Credits / month | USD / month | Best for |
|---|---|---|---|
| Starter (Free) | 2,000 cr | $0 | Testing, hobby projects |
| PAYG Lite | 10,000 cr | $10 | Small creators, indie devs |
| PAYG Pro | 60,000 cr | $50 | Agencies, high-volume apps |
A few things worth noting:
Starter plan — credits are watermarked on audio output, rate-limited to 5 requests/hour, and expire after 30 days. It's a genuine free tier, not a trial.
Pro plan value — Pro gives you 60,000 credits for $50, versus 10,000 for $10 on Lite. That's 6× the credits for 5× the price — a deliberately better deal per credit to reward commitment.
Credit rollover — unused credits roll over for 30 days (Starter), 60 days (Lite), or 90 days (Pro). This matters for teams with variable usage patterns.
Top-Up Packages: When Subscriptions Aren't Enough
If your project needs a one-off burst of synthesis — recording a batch of audio files, narrating a product catalogue, or building out an IVR menu — top-up packages let you buy credits without upgrading your subscription:
| Package | Credits | Rate |
|---|---|---|
| $2 | 2,000 cr | $0.001/cr |
| $5 | 5,000 cr | $0.001/cr |
| $10 | 10,000 cr | $0.001/cr |
| $25 | 25,000 cr | $0.001/cr |
| $50 | 50,000 cr | $0.001/cr |
| $100 | 100,000 cr | $0.001/cr |
The rate is uniform across all packages — $1 always equals 1,000 credits, regardless of package size. No bulk discount tiers to calculate.
Practical Tips for Managing Your Credits
1. Batch your requests when possible. A single 5,000-character request and five 1,000-character requests cost identical credits — but the batched version saves round-trip latency and simplifies your logs.
2. Use character count, not word count, to estimate costs. The average English word is about 5 characters. A 1,000-word article is roughly 5,000 characters → 250 credits → $0.25.
3. Start on the free tier, then upgrade when you hit rate limits. The 5 req/hr limit on Starter will make itself known quickly once you start building real features.
4. Pro Clone is per-character, not per-minute. Once you've trained a Pro Clone voice, generating speech from it costs the same fixed rate regardless of how long your synthesis runs.
5. Pre-calculate before shipping. If you're building a product that will synthesise user-submitted content, run a worst-case character count estimate before launch so your credit balance isn't a surprise.
The Bottom Line
Credit-based pricing for TTS APIs is not complicated once you understand the underlying formula. Most developer workloads cost fractions of a cent per request. The monthly subscription tiers exist for teams that want predictability and want to use voice as a core feature rather than an edge case.
Sign up free and you get 2,000 credits to start — enough to synthesise a short podcast, a product demo, or dozens of test API calls in any of our four Nigerian languages.
9jaLingo Team
Developer Experience · 9jaLingo