Docs / Providers
Doubao Speech (Volcano Engine)
In one sentence: the brand with the most Chinese voices and the richest dialect and character voices. Across its three tiers, the newest 2.0 is actually 40% cheaper than 1.0 — don't be fooled by version numbers.
In one sentence: the brand with the most Chinese voices and the richest dialect and character voices. Across its three tiers, the newest 2.0 is actually 40% cheaper than 1.0 — don't be fooled by version numbers.
What This Is
ByteDance's Volcano Engine speech-synthesis service — the voice tech behind Doubao. Strong Chinese expressiveness, with a voice library full of dialects, characters, and audiobook specialists.
The Three Tiers & Pricing
| Model | Official upstream price | Credits / 1,000 chars | Details |
|---|---|---|---|
| doubao-tts-2.0 | ¥3 / 10,000 chars | 48 | Doubao Speech Synthesis 2.0 |
| doubao-tts-1.0 | ¥5 / 10,000 chars | 80 | LLM Speech Synthesis 1.0 |
| doubao-tts-classic | ¥5.5 / 1,000 calls | ~4 (per call) | Speech Synthesis Small Model |
Price source: https://docs.volcengine.com/docs/6561/1359370 (verified 2026-08-19)
⚠️ 2.0 Is Cheaper Than 1.0, Not More Expensive
This is the easiest thing to get wrong. Intuition says "2.0 is the upgraded 1.0, so it should cost more," but Volcano Engine's pricing runs the other way:
doubao-tts-2.0 → ¥3/10,000 chars → 48 credits/1,000 chars
doubao-tts-1.0 → ¥5/10,000 chars → 80 credits/1,000 chars (67% more)
Default recommendation: 2.0. Only switch to 1.0 if a specific voice you're looking for exists only there — the two versions' voice libraries aren't identical.
The free-preview representative voice Câncân is under 1.0; Soniva auto-switches you to the matching model when you select the Doubao brand.
classic (small model): the Site's Lowest Unit Cost, With a Hard Limit
doubao-tts-classic is Doubao's small-model product line (voice IDs starting with BV).
It's a completely different thing from the first two tiers:
| classic | 1.0 / 2.0 | |
|---|---|---|
| Billing | Per call (¥5.5/1,000 calls) | Per character |
| Per-request length cap | ~280 CJK chars (upstream hard limit of 1024 bytes) | 500 chars |
| Voices | Fixed 24 BV voices | Live-pulled, far more |
| Voice quality | Traditional TTS — steady but flat | LLM voices, natural |
How Cheap Does It Get
Soniva auto-splits long text into segments of ≤280 chars before sending; each segment counts as 1 call and deducts a flat 1 credit:
- Reading 280 chars or fewer → 1 credit
- Reading 1,000 chars → ~4 credits
- Reading 10,000 chars → ~36 credits
Converted, that's ~4 credits/1,000 chars — the lowest unit cost of any cloud voice, even cheaper than Tencent's Premium tier (5).
The trade-off is voice quality — it's traditional TTS, not an LLM voice. It suits "I just need this passage read to me," not "I'm making a finished product."
Voices
1.0 and 2.0 voice lists are pulled live from upstream by Soniva; new upstream voices appear immediately, so this document doesn't hard-code a count — go by what you actually see in the product.
The one exception is classic: the upstream voice-list interface doesn't return BV voices, so this tier uses Soniva's built-in static list, fixed at 24 voices.
Doubao's Chinese dialect voices are its hallmark — like Bay Area Uncle, Adorable Sichuan Girl, Beijing Lad, Guangxi Yuanzhou — hard to find elsewhere.
How to Choose
| Scenario | Pick |
|---|---|
| Mostly Chinese content | 2.0 (48 credits/1,000 chars) |
| Looking for a 1.0-only voice / a dialect | 1.0 (80 credits/1,000 chars) |
| Very long articles, just want it read through, don't mind quality | classic (~4 credits/1,000 chars, the site's most economical) |
| Making finished audio | 2.0 or 1.0 — not classic |
Related Documents
Was this article helpful?
Anonymous feedback is enabled when you allow analytics and the service is available. You can also report an issue by email.
The email includes this article. You can also copy the details and send them to soniva.cde@gmail.com