Terms of Service

These Terms of Service (Terms) are between ARVINDCODEX (Myra, we, us, or our) and the business, developer, or organisation accessing or utilizing the Services (Customer, you, or your).

By accessing myra.arvindcodex.in, calling our APIs, or deploying our voice models, you agree to these Terms. If you accept these Terms on behalf of an enterprise or entity, you represent that you possess full authority to bind that entity.

A signed master enterprise contract, custom service order, or signed data processing agreement controls over these standard online terms in the event of a direct conflict.

Myra provides an enterprise voice-agent and neural text-to-speech platform, including developer APIs, audio streaming WebSockets, fine-tuning tools, and telemetry dashboards (the Services).

Open Source Lineage & Natural Audio Evolution: Our core voice synthesis engine is derived by cloning, fine-tuning, and continuously updating the open-source foundational voice model rumik-ai/rumik-oss-1 (available on Hugging Face). The upstream architecture (combining the Tiny Aya Fire 3B multilingual model with Kyutai Mimi 24 kHz discrete neural speech tokenization) has been substantially upgraded with custom conversational acoustic datasets, expressive inflection tuning, low-latency streaming pipelines, and delivery controls (such as conversational laughs, pauses, sighs, and cadence control) to create hyper-natural, human-grade audio.

We continually upgrade our models to enhance prosody, naturalness, language coverage, and security. We do not materially downgrade core paid functionality during an active subscription term without prior notice.

You are responsible for safeguarding your account credentials, secret API keys, webhook endpoints, and all voice generation activity conducted under your account. You must notify us immediately at contact@arvindcodex.in upon discovering any unauthorized key exposure or account compromise.

You control the text prompts, audio scripts, caller targets, and voice agent behavior. You are solely responsible for ensuring that all audio generation and calling campaigns comply with applicable laws:

• AI Voice Transparency: Provide clear disclosures that the listener is interacting with or hearing an artificial voice, where required by telecommunications, consumer protection, or AI regulations.
• Consent & Authorization: Secure all legally required consents for automated calling, telephony recording, voice analytics, and data processing.
• Do-Not-Call & Regulatory Compliance: Comply with national Do-Not-Call registries, quiet hours, calling limits, and telecommunication carrier rules in every jurisdiction where calls terminate.

Myra does not provide legal advice and does not certify that any specific customer calling campaign or voice script satisfies the legal requirements of any jurisdiction.

You must NOT use the Services or our fine-tuned voice models to:

• Unauthorized Voice Cloning: Synthesize or clone the voice of any real individual without their explicit, documented, and verifiable consent.
• Fraud, Scams & Deception: Generate deepfakes, execute social engineering, spoof caller IDs, or facilitate fraud, extortion, harassment, or defamatory speech.
• Critical High-Risk Workloads: Rely solely on automated speech synthesis in emergency dispatch, critical medical procedures, or high-risk legal/financial adjudications without human oversight.
• Model Weight Extraction: Probe, decompile, extract, or reverse-engineer our fine-tuned model checkpoints or use our APIs to distill or bootstrap a competing commercial model in violation of our terms.
• Malicious Telephony: Send unlawful robocalls, denial-of-service audio floods, or spam to public switched telephone networks.

Customer Content includes all text prompts, custom audio training datasets, scripts, and metadata you upload or submit to the Services.

Ownership of Generated Audio: As between Customer and Myra, Customer retains all right, title, and intellectual property ownership in the specific synthesized audio files, transcripts, and voice outputs generated from Customer Content. Myra claims no copyright or proprietary ownership over the speech audio you generate using your validly licensed account.

Myra does NOT use Customer Content or customer-synthesized audio to train or fine-tune public models without your express written permission.

Open Source Attribution: We acknowledge and credit the open-weights research community. Our voice models are fine-tuned derivatives built upon the foundational work of rumik-ai/rumik-oss-1, Tiny Aya Fire (Cohere For AI), and Kyutai Mimi. We respect open-source licensing conditions (including CC-BY-NC 4.0 upstream research terms and CC-BY-4.0 for the Mimi codec).

ARVINDCODEX Technology Rights: ARVINDCODEX retains all proprietary rights, title, and interest in and to the Myra platform software, orchestration engine, proprietary fine-tuning adapters, latency optimizations, APIs, dashboards, and custom acoustic updates (Myra Technology). No rights in Myra Technology are granted except the limited right to access and use the Services during your subscription.

Data processing is governed by our Privacy Policy. Both parties agree to protect non-public technical, business, and financial information disclosed during the relationship with reasonable care and to use it solely to perform obligations under these Terms.

If you connect the Services to external telephony carriers, SIP trunks, cloud CRMs, or vector databases, you are responsible for compliance with those third-party providers' terms and conditions.

Fees are charged based on audio synthesis duration, character throughput, and concurrent stream limits as set forth in your plan. Fees are non-refundable unless specified otherwise in a signed enterprise order. All prices exclude applicable GST, VAT, and withholding taxes.

Previews, experimental voices, and beta APIs are provided "as is" without service level commitments. We may modify or retire preview features upon reasonable notice.

We may suspend or terminate access immediately if we detect abusive voice generation, non-consensual voice cloning, illegal scam campaigns, security breaches, or non-payment. Either party may terminate with 30 days' written notice for material uncured breach.

Myra warrants that the Services will perform substantially in accordance with our documentation.

Except as expressly provided, the Services, models, and audio outputs are provided "as is" without warranty of any kind, whether express, statutory, or implied, including merchantability, fitness for a particular purpose, or non-infringement.

Neural voice synthesis is generative and probabilistic. You are responsible for validating audio output appropriateness before deploying it in live recipient-facing environments.

You agree to defend, indemnify, and hold harmless ARVINDCODEX and its officers, engineers, and affiliates from any third-party claims, penalties, or damages arising out of your Customer Content, unauthorized voice cloning, violations of telecommunications laws, or breach of these Terms.

To the maximum extent permitted by law, neither party will be liable for indirect, incidental, special, consequential, or punitive damages, or loss of profits or revenue.

Each party's total cumulative liability arising out of or relating to these Terms will not exceed the amounts actually paid by Customer to Myra in the twelve (12) months preceding the incident, or USD 100 if no fees were paid.

These Terms are governed by and construed in accordance with the laws of India. Any legal action or dispute arising under these Terms shall be subject to the exclusive jurisdiction of the competent courts in India.

These Terms constitute the entire agreement between the parties regarding the subject matter herein. If any provision is deemed unenforceable, the remaining provisions will remain in full force.

Official notices or inquiries should be sent to:
ARVINDCODEX — Myra Legal Operations
Website: myra.arvindcodex.in
Email: contact@arvindcodex.in