Loading page…
Bodhi TTS gives businesses tools to correct names and numbers without retraining. Its speed, deployment and rival-price comparisons remain company claims.
Listen to this story
Navana.ai’s Bodhi TTS is now available for businesses seeking Indian-language speech generation, with a list price of ₹12 per 10,000 characters and more than 50 voices. Its commercial pitch centers on operational control: teams can correct pronunciations through dictionaries and optionally deploy the model in their own data center. Navana.ai claims audio begins in under 100 milliseconds, but that is first-audio latency—not full response time.
Coverage spans Hindi, Telugu, Tamil, Marathi, Bengali, Odia, Malayalam, Kannada, English and Gujarati.
Navana.ai says teams can upload name dictionaries and correct pronunciations without retraining; voice cloning uses a few seconds of reference audio.
The company compares ₹12 per 10,000 characters with ₹30 for Sarvam AI’s Bulbul v3 and about ₹95 for ElevenLabs; these are vendor-supplied, not independently normalized estimates.
A bank’s automated caller can sound fluent and still get a customer’s name or loan details wrong. Navana.ai has launched Bodhi TTS, a model that turns text into speech across 10 languages, with controls for how particular words are pronounced. The company lists it at ₹12 per 10,000 characters, pitching a way for businesses to make large volumes of customer calls without paying separately for custom voices.
The release covers Hindi, Telugu, Tamil, Marathi, Bengali, Odia, Malayalam, Kannada, English and Gujarati, with more than 50 voices. Navana.ai is aiming the model at real-time applications: it says speech starts in under 100 milliseconds, while the rest of the sentence is still being generated. That measures the start of audio, not how long a complete call response takes.
Navana.ai says Bodhi TTS is designed to speak currency amounts, phone numbers, PAN numbers, dates, loan installments, scheme names and account numbers in forms familiar to Indian businesses. A company can upload a dictionary of product, branch and customer names to steer pronunciation. It says a correction does not require retraining the model, which would let a team address a name that comes out wrong without rebuilding its speech system.
The other customization tool is voice cloning. Navana.ai says a developer can submit a few seconds of reference audio through its API and generate a custom voice without additional training. It says the reference is cached so that using the voice adds no delay during a call. Whether that holds across customer deployments is a different question from whether the feature is available.
Navana.ai lists Bodhi TTS at ₹12 per 10,000 characters, with all voices and features included.
Navana.ai cites ₹30 per 10,000 characters for Sarvam AI’s Bulbul v3. This rival-price figure comes from Navana.ai’s comparison.
Navana.ai puts comparable ElevenLabs pricing at about ₹95 per 10,000 characters. This is its own comparison, not a verified like-for-like service-cost estimate.
Navana.ai says every Bodhi TTS voice and feature is covered by that list price, rather than a premium tier. It estimates the charge at about ₹0.72 per minute of generated speech. The rival figures above are also supplied by Navana.ai; a buyer comparing services would still need to weigh voice quality and the amount of speech its own calls generate. The stated character rate is a clearer starting point than a projected per-call bill.
For a bank deciding where customer information goes, Navana.ai makes a separate deployment claim. It says Bodhi TTS can run inside an enterprise’s own data center, keeping audio, transcripts and personal data within that organization’s security environment. The company says the model is roughly five times smaller than the nearest comparable architecture, but does not identify that architecture in its comparison. An in-house installation is an option it describes, not a guarantee about every customer setup.
Bodhi TTS is available through Navana.ai’s platform, with ₹1,000 in signup credits for new users. The start date is less tidy than the access status: Moneycontrol says it was available from September 23, while a company-supplied release published by ANI says public access opened September 24. Neither account changes the central development: customers can now try the model rather than wait for a planned release.
The next useful evidence will come from the calls Bodhi TTS is meant to serve: whether its voices handle unfamiliar names and financial details clearly, and whether fast first audio leads to a responsive exchange. The launch supplies features, a price and Navana.ai’s performance claims, but not results from customers testing those claims across their own calls. Voice cloning raises another practical choice: when, if ever, should a business use a copied voice to speak with a customer?
Loading discussion...
Join the conversation
What would make that feel useful or risky?
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.