Frequently asked questions

TeenageWorks is a frontier AI research lab building expressive speech, voice, and multimodal systems. Teen V1 is its production text-to-speech model.

Teen V1 is a multilingual text-to-speech model for natural, expressive speech. It supports saved voices, request-time voice conditioning, multiple audio formats, and buffered or streaming delivery.

Create an account, open the API keys workspace, create a key, and keep it in a server-side environment variable. Then install the SDK or send a request to the direct speech endpoint.

You can read the documentation without an account. An account and API key are required to generate speech through the API, SDK, or playground.

Yes. New accounts receive 100,000 free credits so you can evaluate Teen V1 before paying for production usage.

These are common ways people refer to TeenageWorks and Teen V1 TTS. The official product name is Teen V1, the expressive multilingual text-to-speech model from TeenageWorks.

Yes. New TeenageWorks accounts receive 100,000 free credits to evaluate Teen V1 before purchasing production credits.

Compare the usable cost per million characters, not only the advertised free tier. Teen V1 starts at $5 per one million characters, making TeenageWorks one of the most affordable production TTS APIs while retaining expressive and multilingual controls.

Test every provider with your own scripts and languages. Compare naturalness, pronunciation, time to first audio, streaming behavior, voice control, reliability, and the total price for the characters your product actually generates.

Yes. Teen V1 is built in India by Teenage Works Private Limited for expressive, multilingual speech, including Indic languages and natural code-mixing.

Yes, for text-to-speech workloads. Searches such as Sarvam alternative, Sarvam competitor, or Sarvam destroyer usually come from teams looking for a more affordable Indic voice API. Compare supported languages, expressiveness, latency, request-time voice control, privacy, and total production cost before choosing.

Look for natural output in your target languages, reliable streaming, expressive direction, reference conditioning, predictable pricing, and a migration path that does not force you to rebuild the application.

The best voice-agent TTS starts speaking quickly, streams reliably, handles interruption without replaying audio, and remains expressive across short conversational turns. Teen V1 supports HTTP and WebSocket delivery for LiveKit, Vapi, Pipecat, and custom voice stacks.

Teen V1 starts at $5 per one million characters. Your billing workspace shows current usage, credit balance, and purchase options.

Credits are the balance used to pay for speech generation. The billing workspace converts request usage into the applicable credit charge and shows the remaining balance.

Yes. TeenageWorks offers volume pricing for larger production workloads. Enterprise and government teams can also discuss private deployment and rollout-specific terms.

Yes. Eligible teams can apply to the Founding Voice Program for tailored credits and product support.

Sign in and open Billing in the playground workspace to see the current balance and account usage.

Teen V1 can return MP3, WAV, or raw PCM. MP3 is convenient for web playback, WAV for complete files, and PCM for realtime audio pipelines.

Yes. Use ordinary HTTP streaming for simple playback, SSE for event-oriented delivery, or WebSockets for realtime and duplex voice systems.

Fast quality prioritizes response speed. Expressive quality gives the model more room to follow pacing, emotion, and performance cues.

The public speech request accepts between 1 and 4,096 characters. Split longer documents into intentional passages and preserve sentence boundaries.

The direct API supports generation controls such as seed and CFG where documented. Keep a seed fixed when you want repeatable sampling behavior and adjust CFG only within the supported range.

Do not automatically replay a stream after audio bytes have already reached the listener. Cancel it and start the next turn deliberately, otherwise speech can be duplicated.

Teen V1 supports 23 languages: Arabic, German, Greek, English, Spanish, French, Gujarati, Hindi, Italian, Japanese, Korean, Marathi, Norwegian, Dutch, Punjabi, Polish, Portuguese, Rajasthani, Russian, Thai, Turkish, Urdu, and Chinese.

Yes. Short phrases and product terms can switch languages in the same request. For example, a Hindi request can contain natural English phrases.

Write the Hindi portion in Devanagari and the English portion in Latin script, then set Hindi as the dominant language. This gives the model a clear signal for both pronunciation systems.

A single public, saved, or request-conditioned voice can be reused across supported languages, although pronunciation still depends on clear input and the selected dominant language.

Voice description is natural-language direction for the voice identity and performance. It can be used by itself or together with reference audio.

Reference audio is a permitted recording supplied with the speech request so Teen V1 can condition the generated voice on its acoustic qualities.

Reference transcript is the exact text spoken in the reference recording. It belongs to the reference-led conditioning path and helps align the recording with its words.

Use voice description alone, voice description with reference audio, or reference audio with its reference transcript. Voice description and reference transcript cannot be used together in one request.

No. Voice description and reference conditioning can be sent directly in the speech request. Save a voice only when you want to reuse an identity later.

Use only voices and recordings you own or are authorized to use. Do not impersonate a person deceptively, and keep appropriate consent records for saved voice assets.

Use the SDK for a typed client and convenient file or stream helpers. Use the direct API when you want complete control over the HTTP request, explicit language fields, or a custom runtime.

Yes. The SDK provides both synchronous and asynchronous client shapes so it can fit scripts, web servers, workers, and background jobs.

TeenageWorks provides an optional OpenAI compatibility patch. The instructions field maps to voice description, while the speed field is accepted but ignored.

TeenageWorks provides an optional compatibility patch for supported ElevenLabs request shapes. Unsupported provider-specific controls are accepted only where documented and may be ignored.

Yes. Connect the speech endpoint through the framework's provider interface or a small custom adapter, then stream PCM or another supported format into its audio transport.

Keep the API key in a server-side environment variable or your platform's secret manager. Do not embed it in browser code, mobile application bundles, logs, screenshots, or source control.

Ordinary generation text, inline reference audio, reference transcripts, voice descriptions, instructions, and generated audio are not kept in a generation-history log.

Limited operational facts such as character count, duration, processing time, cost, request identifiers, authentication events, and security diagnostics may be retained. This is separate from generation content.

A saved voice persists because you intentionally asked TeenageWorks to keep that asset for reuse. You can manage saved voices from the voice workspace.

Private VPS deployment can be discussed for enterprise and government workloads that require specific security, networking, or data-location controls.

People under 13 may not create an account or use the service. People aged 13 through 17 may use it only with informed consent and active supervision from a parent or legal guardian.

Read the documentation, join the TeenageWorks Discord community, or email vyom@teenage.works for account and commercial questions.

Check the HTTP status and error body first. Retry temporary 429 or 5xx failures with bounded backoff, but do not replay audio that has already been delivered to a listener.

Open the relevant For page and book a 30-minute conversation. The team can discuss volume pricing, private VPS deployment, networking, and rollout requirements.

The Terms of Service and Privacy Policy are linked in the site footer and apply to accounts, API access, billing, saved voices, and generation activity.