Billing
Per character
Spaces, newlines and most SSML markup count toward billing.
Google Cloud Text-to-Speech provider review
Google TTS: Voices, Pricing and Real-World Performance
Google Cloud TTS is fast and predictable, but the voice family matters more than the brand name. Compare Standard, WaveNet, Neural2, Chirp 3 HD and Studio pricing, controls and production behavior before choosing a voice.
Pricing and model facts checked July 16, 2026Billing: Per characterEntry price: $4 / 1MProduction snapshot: 578 ms median
Quick verdict
Google Cloud Text-to-Speech is a strong default for applications that generate many short or repeatable speech requests. Standard voices remain unusually useful when cost, concurrency and predictable output matter. Premium families can improve quality, but each model should be tested for punctuation, language and SSML compatibility instead of being selected only because it is more expensive.
Strong fit
- ✓High-volume text, subtitle and multi-voice workloads
- ✓Applications that need character-based pricing and predictable APIs
- ✓Mixed-language scripts with product names or English terms
- ✓Teams that want SSML control for pauses, numbers and pronunciation
Trade-offs to test
- !Quality and control vary substantially by voice family
- !Some newer voices do not support the same SSML features as legacy voices
- !The most expensive voice is not automatically the best for short subtitle cues
- !Free usage and price differ by model family
Entry price
$4 / 1M
Current Standard and WaveNet price after the free monthly allowance.
Production snapshot
578 ms median
50 successful short Standard requests from internal logs on July 15, 2026.
Control
Voice + SSML
Pitch, rate, breaks, say-as and pronunciation depend on voice support.
What is Google Cloud Text-to-Speech?
Google Cloud Text-to-Speech is the traditional Google Cloud speech synthesis API. You choose a language, an exact voice ID, an audio format and optional SSML or audio settings. This makes it different from Gemini TTS, where delivery is directed mainly with natural-language prompts.
Google TTS is not one quality level. Standard and WaveNet prioritize efficient, repeatable generation; Neural2 and Polyglot occupy a higher-priced neural tier; Chirp 3 HD targets newer high-definition output; Studio is a premium family for supported narration scenarios. The correct choice depends on language, cue length, throughput and the exact controls your workflow needs.
In TTS For Free, Google voices are commonly used for text-to-speech, subtitle cues and multi-voice scripts because many small requests can finish quickly. Our user behavior also shows that Standard voices continue to be selected frequently even when premium options are available.
Google TTS voice families
This table separates current official pricing from our recommended test hypothesis. Voice availability and feature support still need to be checked by exact locale and voice ID.
| Voice family | Free monthly usage | Price after free usage | Control notes | Best-fit starting point |
|---|---|---|---|---|
| Standard | 4M characters | $4 / 1M characters | Broad legacy SSML workflow | High-volume reading and subtitle cues |
| WaveNet | 4M characters | $4 / 1M characters | Check exact voice availability | General neural speech at low cost |
| Neural2 | 1M characters | $16 / 1M characters | Feature support varies by voice | Narration where naturalness matters |
| Polyglot (Preview) | 1M characters | $16 / 1M characters | Preview and locale-specific | Cross-language tests |
| Chirp 3: HD | 1M characters | $30 / 1M characters | Test SSML and punctuation compatibility | Premium natural speech after validation |
| Studio | 1M characters | $160 / 1M characters | Limited supported voice/language scenarios | Premium studio narration |
Official Google Cloud pricing checked July 16, 2026. Google can change model availability and pricing; verify again before publishing a long-lived cost estimate.
SSML, pronunciation and mixed-language scripts
Google SSML is useful when plain punctuation is not enough. It can add pauses, control prosody, spell identifiers and provide pronunciation guidance. TTS For Free also has a dedicated Google SSML editor for users who need structured markup rather than a simple text box.
Mixed-language text is a practical strength in our workload. Vietnamese sentences containing English product names often remain understandable without splitting every language into a separate request. Results still vary by voice, so names, acronyms, URLs, dates and currencies should be included in the acceptance test.
Read an identifier clearly
<speak> Your order number is <say-as interpret-as="characters">A17B</say-as>. </speak>
Add a controlled pause
<speak> The first result is ready. <break time="400ms"/> Review it before generating the full batch. </speak>
Mixed Vietnamese and English test
TTS For Free hỗ trợ Google Cloud Text-to-Speech cho video YouTube, khóa học online và workflow SRT to Speech.
What to verify
Test numbers, acronyms, brand names, punctuation, <lang> behavior and whether the selected voice accepts the SSML tags you plan to use.
How much does Google TTS cost?
Google charges for the number of characters submitted, including spaces and newlines. Most SSML tags are billable too. Cost examples below exclude retries and other Google Cloud services.
| Scenario | Standard / WaveNet | Neural2 / Polyglot | Chirp 3 HD | Studio |
|---|---|---|---|---|
| 10,000 characters | $0.04 | $0.16 | $0.30 | $1.60 |
| 100,000 characters | $0.40 | $1.60 | $3.00 | $16.00 |
| 1M characters | $4.00 | $16.00 | $30.00 | $160.00 |
| 1 hour at 50k–70k chars | $0.20–$0.28 | $0.80–$1.12 | $1.50–$2.10 | $8.00–$11.20 |
The one-hour row uses an editorial assumption of 50,000–70,000 characters per spoken hour. Actual speaking rate and language can change the result.
What we observed in production
The snapshot below comes from 50 successful Google Cloud Standard request records in TTS For Free logs on July 15, 2026. The sample contains short English and Vietnamese requests, so it describes this workload—not every Google voice family.
Successful records
50
This snapshot contains completed requests only.
Input range
14–69 chars
Short UI and subtitle-like lines.
Median time
578 ms
Observed request processing time.
P95 / range
853 ms
Observed range: 305–922 ms.
Short Standard requests were consistently fast in this captured sample, which matches why Google is frequently selected for SRT-to-speech and multi-voice workloads with many small lines.
Standard voices remain popular with users even when higher-priced families are available. For production, clarity, timing and retry behavior can matter more than the model label.
Vietnamese lines containing English terms are worth testing on Google before splitting them into separate language requests; splitting can reduce context and create unnatural transitions.
Chirp 3 HD and Studio should be evaluated with the exact punctuation and text-splitting strategy used by the application. Premium pricing does not remove input-preparation problems.
This is an internal operational snapshot, not a controlled provider benchmark. It does not include failed requests, network decomposition, concurrency level or premium voice families. Do not compare it directly with a different provider sample that uses longer inputs.
When Google TTS is a good choice
Many short speech requests
Subtitle cues, UI responses and role-based scripts benefit from fast, repeatable generation.
Cost-sensitive document reading
Standard and WaveNet character pricing can be efficient for large amounts of straightforward narration.
SSML-driven applications
Use structured control when numbers, pauses, pronunciation or speaking rate must be predictable.
When to test another option first
Highly directed acting
Gemini, OpenAI, Azure styles or creator-focused models may provide easier emotion direction.
Voice cloning is essential
Use a provider and product designed around consented custom or cloned voices.
You selected a model only because it says HD
Run a same-script test for punctuation, timing, language and cost before committing.
Google TTS compared with other providers
The useful comparison is not which provider wins overall, but which control and billing model matches the workload.
| Criteria | Google Cloud TTS | Azure TTS | Gemini TTS | OpenAI TTS | ElevenLabs |
|---|---|---|---|---|---|
| Main control | Voice + SSML | SSML + supported styles | Natural-language prompt | Natural-language instructions | Voice, model, settings and tags |
| Pricing unit | Characters by family | Characters / tier | Text and audio tokens | Model-specific tokens | Credits / usage / plan |
| Operational fit | Fast repeatable synthesis | Expressive enterprise speech | Creative prompted speech | Conversational prompted speech | Creator and brand voice workflows |
| Subtitle chunks | Strong starting point; benchmark exact voice | Good; style support is voice-specific | Test consistency and long generation | Test rate limits and chunk consistency | Test plan concurrency and regeneration cost |
Try Google voices with the same script
The tool below is filtered to Google voices available in TTS For Free. Start with a short representative script, then test the exact language, punctuation and SSML needed by your project.
Limit exceeded by 103 chars! Resets at .
🔊 You are using standard Text-to-Speech.Free users have daily character limits and limited premium voices. Upgrade for premium voices, longer text, and better stability.
103/2000 charactersLogin for higher limits
Pick language, voice, and generate audio.
Select Language
Voice Settings
No items
Tutorial Video
Continue the Google TTS workflow
Google SSML Editor
Build and test pauses, say-as, prosody and pronunciation markup.
Open pageSRT to Speech
Generate timed speech from subtitle files with Google or another supported provider.
Open pageMulti-Voice TTS
Assign voices to multiple roles in the same script.
Open pageGemini TTS
Compare traditional Google Cloud synthesis with prompt-controlled Gemini speech.
Open pageAI Dubbing Workflow
See how transcription, translation, voice mapping and QA fit together.
Open pageOfficial sources
Frequently Asked Questions
Q: Is Google TTS free?
A: Google Cloud currently includes monthly free usage that varies by voice family. Billing must still be enabled, and usage above the allowance is charged.
Q: Which Google TTS voice family should I choose?
A: Start with Standard or WaveNet for cost and repeatability, Neural2 for a higher neural tier, and test Chirp 3 HD or Studio only after confirming language, control and pricing fit.
Q: Does Google TTS support SSML?
A: Yes, Google Cloud TTS supports SSML features such as breaks, say-as and prosody, but exact support can vary by voice and model.
Q: Is Google TTS good for Vietnamese?
A: It can work well for Vietnamese and mixed Vietnamese-English scripts, but pronunciation should be tested with the exact voice, names and terminology in your content.
Q: Is Chirp 3 HD always better than Standard?
A: No. Chirp 3 HD is newer and more expensive, but Standard may still be a better operational fit for short subtitle cues, cost-sensitive work or SSML-dependent workflows.
Q: How fast is Google TTS?
A: In one internal snapshot of 50 successful short Standard requests, median processing time was about 578 ms and p95 was 853 ms. This is not a universal provider benchmark.