Freedom251 › Text-to-speech tools
Spex-TTS
What Spex-TTS is
Spex-TTS is a text-to-speech engine available through web and API access, with voice cloning in 23 languages. It grew from a production live-translation product. The API can return complete WAV audio, raw PCM, HTTP server-sent event streams, or WebSocket audio streams. Numeric settings and inline transcript tags shape delivery, including timed breaks and spelled-out text. The maker reports first audio byte in 146 ms over the open internet and 116 ms p50 when the server is warm. It reports a 1.98% word error rate against a 2.17% human reference using the same scorer. Customer text and reference audio are excluded from training according to the maker. Generated samples carry an inaudible watermark that survives transcoding and cannot be disabled through configuration. Users must obtain speaker consent, although the service does not detect it. Usage is auditable by API key. Plans include a free tier and paid options from $10 a month.
Who it suits
Spex-TTS suits developers needing API-based speech generation, streaming audio, or voice cloning across 23 languages. It also fits users who need web access and documented privacy and infrastructure claims.
What it does well
- Voice cloning across 23 languages
- WAV, raw PCM, SSE, and WebSocket output
- Expression controls for pauses and spelling
- Free plan with 50,000 starting tokens
- Usage auditing by API key
What to know first
- No voice-management API
- Speaker consent is not detected
- Generated audio always includes an inaudible watermark
- No free trial
Plans
| Plan | Billed and limits | Price |
|---|---|---|
| Free | 50,000 tokens to start | Free |
| Take a Gandr | Resets monthly · 1 million tokens a month | $10 a month |
| Pro | Resets monthly · 5 million tokens a month | $50 a month |
| Scale / flat rate | $150 per stream, per month, on annual; $180 month-to-month · One unlimited stream · Unmetered | $150 |
| Enterprise | Priced together · 500 lines and up · Volume, burst and failover | On request |
As gandr.ai lists them, read 9 Oct 2026. The maker's own page has the final word.
Verdict
Spex-TTS combines multilingual voice cloning with several API streaming formats and detailed expression controls. Its free plan provides an entry point, while consent handling and the missing voice-management API are important constraints.
Facts from Spex-TTS's own pages
- Accuracy
- The maker reports a 1.98% word error rate against a 2.17% human reference using the same scorer. · gandr.ai, 9 Oct 2026
- API access
- Yes · gandr.ai, 20 Sept 2026
- API limits
- The maker says there is no voice-management API; a voice reference is sent in the request, and the API example uses a ten-second recording. · gandr.ai, 9 Oct 2026
- Consent limit
- The maker requires users to have the reference speaker’s consent but says consent is not detected by the service. · gandr.ai, 9 Oct 2026
- Export formats
- ["WAV","raw PCM (pcm_s16le)"] · gandr.ai, 20 Sept 2026
- Expression controls
- Numeric controls and inline transcript tags can steer speech expression, including timed breaks and spelled-out text. · gandr.ai, 9 Oct 2026
- Free plan
- Yes · gandr.ai, 20 Sept 2026
- Integrations
- The maker documents API access through REST, server-sent events, and WebSocket; no named third-party integrations are listed on the opened API page. · gandr.ai, 9 Oct 2026
- Languages
- 23 · gandr.ai, 20 Sept 2026
- Languages
- The maker says one cloned reference can be used across 23 languages. · gandr.ai, 9 Oct 2026
- Performance
- The maker reports first audio byte in 146 ms over the open internet and 116 ms p50 when the server is warm. · gandr.ai, 9 Oct 2026
- Platforms
- ["web","api"] · gandr.ai, 20 Sept 2026
- Privacy
- Customer text and reference audio never enter a training set, according to the maker. · gandr.ai, 9 Oct 2026
- Purpose
- Spex-TTS is a text-to-speech engine that grew out of a production live-translation product. · gandr.ai, 9 Oct 2026
- Security claims
- The maker’s site lists SOC 2 Type 2 infrastructure, HIPAA-Verified infrastructure, and GDPR-Ready infrastructure. · gandr.ai, 9 Oct 2026
- Paid from
- 10 · gandr.ai, 20 Sept 2026
- Streaming
- The API offers complete WAV responses, HTTP server-sent event streams, and WebSocket audio streams. · gandr.ai, 9 Oct 2026
- Synthetic audio marking
- Every generated sample carries an inaudible watermark that the maker says survives transcoding and cannot be skipped by configuration. · gandr.ai, 9 Oct 2026
- Usage tracking
- Usage is auditable per API key through GET /v1/usage. · gandr.ai, 9 Oct 2026
- Voice cloning
- The API supports instant voice cloning from ten seconds of a consenting speaker’s audio, and the clone speaks 23 languages. · gandr.ai, 9 Oct 2026
Where Spex-TTS stands on Freedom251
- Text-to-speech tools#67 of 232
- Text-to-Speech Software#126 of 223