Inworld AI Releases TTS-1.5 For Realtime, Production Grade Voice Agents

Inworld AI has introduced Inworld TTS-1.5, an upgrade to its TTS-1 family that targets realtime voice agents with strict constraints on latency, quality, and cost. TTS-1.5 is described as the number top ranked text to speech system on Artificial Analysis and is designed to be more expressive and more stable than prior generations while remaining suitable for large scale consumer deployments.

Realtime latency for interactive agents

TTS-1.5 focuses on P90 time to first audio latency, which is a critical metric for user perceived responsiveness. For TTS-1.5 Max, P90 time to first audio is below 250 ms. For TTS-1.5 Mini, P90 time to first audio is below 130 ms. These values are about 4 times faster than the prior TTS generation according to Inworld.

The TTS-1.5 stack supports streaming over WebSocket so synthesis and playback can start as soon as the first audio chunk is generated. In practice this keeps end to end interaction latency in the same range as typical realtime language model responses when models run on modern GPUs, which is important when TTS is part of a full agent pipeline.

Inworld recommends TTS-1.5 Max for most applications because it balances latency near 200 ms with higher stability and quality. TTS-1.5 Mini is positioned for latency sensitive workloads such as real time gaming or ultra responsive voice agents where every millisecond is important.

Expression, stability and benchmark position

TTS-1.5 builds on TTS-1 and it delivers about 30 percent more expressive range and about 40 percent better stability than the earlier models.

Here expression refers to features such as prosody, emphasis, and emotional variation. Stability is measured by metrics such as word error rate and output consistency across long sequences and varied prompts. The reduction in word error rate reduces issues like truncated sentences, unintended word substitutions, or artifacts, which is important when TTS output is driven directly from generated language model text.

Pricing and cost profile at consumer scale

TTS-1.5 is priced with two main configurations. Inworld TTS-1.5 Mini costs 5 dollars per 1 million characters, which is about 0.005 dollars per minute of speech. TTS-1.5 Max costs 10 dollars per 1 million characters, which is about 0.01 dollars per minute.

This cost profile makes it feasible to run TTS continuously in high usage products such as voice native companions, education platforms, or customer support lines without TTS becoming the dominant variable cost.

Multilingual support, voice cloning and deployment options

Inworld TTS-1.5 supports 15 languages. The list includes English, Spanish, French, Korean, Dutch, Chinese, German, Italian, Japanese, Polish, Portuguese, Russian, Hindi, Arabic, and Hebrew. This allows a single TTS pipeline to cover a wide set of markets without separate models per region.

The system provides instant voice cloning and professional voice cloning. Instant voice cloning can create a custom voice from about 15 seconds of audio and is exposed directly in the Inworld portal and through API. Professional voice cloning uses at least 30 minutes of clean audio, with 20 minutes or more recommended for best results, and targets branded voices and less common accents.

For deployment, TTS-1.5 is available as a cloud API and also as an on prem solution, where the full model runs inside the customer infrastructure for data sovereignty and compliance. The same quality profile is maintained across both deployment modes, and the models integrate with partner platforms such as LiveKit, Pipecat, and Vapi for end to end voice agent stacks.

Key Takeaways

• Inworld TTS 1.5 delivers realtime performance, with P90 time to first audio under 250 ms for the Max model and under 130 ms for the Mini model, about 4 times faster than the prior generation.
• The model increases expressiveness by about 30 percent and improves stability with about 40 percent lower word error rate.
• Pricing is optimized for consumer scale, TTS 1.5 Mini costs about 5 dollars per 1 million characters and TTS 1.5 Max costs about 10 dollars per 1 million characters, which is significantly cheaper per minute than many competing systems.
• TTS 1.5 supports 15 languages and offers instant and professional voice cloning, enabling custom and branded voices from short reference audio or longer recorded datasets.
• The system is available as a cloud API and as an on prem deployment, and integrates with existing voice agent stacks, which makes it suitable for production realtime agents that require explicit guarantees on latency, quality, and data control.

Check out the Technical details. Also, feel free to follow us on Twitter and don’t forget to join our 100k+ ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.