Five-language generation
Choose Chinese, English, Japanese, Spanish or Arabic explicitly for each generation.
IndexTTS 2.5
Use the latest IndexTTS model with a short reference voice, explicit language selection, emotion controls and speaking-pace presets across Chinese, English, Japanese, Spanish and Arabic.
IndexTTS Online Voice Cloning
OnlineAdd a short reference voice and the text you want it to speak.
Choose the IndexTTS model for this generation.
Add 3–30 seconds of clear, single-speaker audio.
Choose the language of the text for IndexTTS 2.5.
Add a reference voice and text to get started.
WAV output · Signed-in generations can be viewed in History.
By generating, you confirm that you are 18 or older and that you own this reference voice or have permission to use it. Use is subject to our Terms and Acceptable Use Policy.
API early access
Tell us what you are building, expected audio volume and timing. We will use real demand to prioritize hosted API access.
IndexTTS model samples
Use official source samples to compare IndexTTS2 with the latest multilingual IndexTTS 2.5 model before choosing a model for generation.
IndexTTS 2.5 is available online with Chinese, English, Japanese, Spanish and Arabic generation.
Choose a sample
What is new
The 2.5 path keeps the same reference-voice workflow while adding multilingual controls that matter for localization and repeat generation.
Choose Chinese, English, Japanese, Spanish or Arabic explicitly for each generation.
Follow the reference delivery or use the supported Happy, Sad, Angry and Calm presets.
Use Faster, Normal or Slower presets without exposing raw inference parameters.
Saved Voices, History, downloads, quotas and billing remain shared with IndexTTS2.
Languages & controls
IndexTTS 2.5 exposes a language selector because language changes the generation request. The first public UI keeps advanced pronunciation and raw sampling controls hidden.
Model comparison
Choose 2.5 when you need multilingual generation or pace control. Choose IndexTTS2 when the established English/Chinese path fits the job. Neither model is treated as a separate product.
IndexTTS2Reuse a reference voice across supported languages for localized media and creator workflows.
Generate narration in Japanese, Spanish or Arabic without moving to a separate product surface.
Keep a reusable voice asset while changing the target language and delivery controls per generation.
Upload or record a short, clear single-speaker reference, or choose a saved voice after signing in.
Enter the text, select one of the five supported languages, choose emotion and set Faster, Normal or Slower pace.
Generate the speech, listen to the result, retry if needed and download WAV output.
IndexTTS Online is an independent third-party hosted service. Use the official project pages for upstream model research and source information.
Yes. The official IndexTTS project released IndexTTS 2.5 in August 2026.
Chinese, English, Japanese, Spanish and Arabic are supported in the 2.5 generation path.
Yes. Saved Voices are model-independent reference assets and can be reused directly with the IndexTTS 2.5 workflow.
No. Both models remain part of IndexTTS Online and share the same account, voice library, History and pricing system.
Yes. The hosted 2.5 UI exposes Faster, Normal and Slower pace presets.
Hosted workflow
The official IndexTTS project remains the source for upstream research, code and model releases. IndexTTS Online is an independent hosted product for people who want a repeatable browser workflow around that model.
Upload or record a reference voice and generate with IndexTTS 2.5 without setting up the model locally.
Saved Voices, History and WAV downloads support repeat use instead of a one-off model test.
Use IndexTTS2 and IndexTTS 2.5 inside the same account, voice library and product flow.
Developer demand
The browser-hosted IndexTTS 2.5 workflow is live. A public self-serve API is not open today. If hosted API access would be useful for your product or automation, record your interest so we can measure real developer demand before building it.
Submitting this form does not create an API account. We may contact you by email about an early pilot.
Practical guides
Use focused guides to improve reference audio, choose models, prepare scripts, and compare results under controlled conditions.
A practical multilingual voice-cloning workflow covering script preparation, pronunciation checks, brand terms, review, and consistency across languages.
Read guideA practical comparison of IndexTTS2 and IndexTTS 2.5 based on language coverage, controls, workflow needs, and production trade-offs.
Read guidePractical guidance for using the shared hosted emotion presets and IndexTTS 2.5 pace controls while preserving intelligibility, consistency, and natural delivery.
Read guidePractical Japanese script-preparation guidance for sentence boundaries, punctuation, readings, counters, loanwords, honorific tone, and review before multilingual voice generation.
Read guidePrepare Spanish scripts for synthetic speech with better punctuation, numbers, abbreviations, regional vocabulary, names, questions, and multilingual review.
Read guidePrepare Arabic text for synthetic speech with better script direction, diacritics where needed, numbers, names, punctuation, Modern Standard Arabic choices, and human review.
Read guide