Three new Kitten TTS models – smallest less than 25MB
Kitten TTS (https://github.com/KittenML/KittenTTS) is an open-source series of tiny and expressive text-to-speech models for on-device applications. We had a thread last year here: https://news.ycombinator.com/item?id=44807868.Today we're releasing three new models with 80M, 40M and 14M parameters.The largest model (80M) has the highest quality. The 14M variant reaches new SOTA in expressivity among similar sized models, despite being <25MB in size. This release is a major upgrade from the previous one and supports English text-to-speech applications in eight voices: four male and four female.Here's a short demo: https://www.youtube.com/watch?v=ge3u5qblqZA.Most models are quantized to int8 + fp16, and they use ONNX for runtime. Our models are designed to run anywhere eg. raspberry pi, low-end smartphones, wearables, browsers etc. No GPU required! This release aims to bridge the gap between on-device and cloud models for tts applications. Multi-lingual model release is coming soon.On-device AI is bottlenecked by one thing: a lack of tiny models that actually perform. Our goal is to open-source more models to run production-ready voice agents and apps entirely on-device.We would love your feedback!
FL score
out of 100
Verdict
high confidence
Competition
9
competitors found, emerging market, funded players
Trend
No signal yet
Build SOTA tiny, expressive, on-device TTS models (<25MB, no GPU) to enable private, low-latency voice applications, monetizing via commercial licenses and enterprise support.
The pain
The gap
Build angle
Strengths
- Addresses a clear, severe, and growing pain point in the on-device AI market.
- Strong value proposition: SOTA expressivity in <25MB models, no GPU required, runs anywhere.
- Directly solves critical issues of cloud TTS (latency, cost, privacy).
- Capitalizes on the unreliability and installation pain of a major open-source competitor (Coqui TTS).
- High demand from a specific, reachable developer audience.
- Open-source nature provides excellent distribution and leverage.
Risks
- Sustaining SOTA ML model development as a solo builder is technically challenging and time-consuming.
- Converting open-source usage into a viable monetization strategy (e.g., commercial licenses, enterprise support) requires business acumen and marketing.
- Competition from other small TTS models (e.g., Piper, Kokoro) could intensify.
- Ensuring robust, production-ready quality and broad device compatibility for all supported platforms.
Next steps
- Focus on developer experience: make installation and integration exceptionally easy, especially compared to Coqui TTS.
- Actively engage with developers on GitHub, Hacker News, and relevant forums to gather feedback and build a community around the new models.
- Explore and validate potential commercialization paths (e.g., early enterprise partnerships, commercial licenses) alongside the open-source release.
- Document performance benchmarks rigorously, particularly against competitors in the 'tiny model, high expressivity, no GPU' niche.
Fly Labs Method
Is the pain real, is there a gap, is it the right time, can one person build it.
- Problem clarity
- 26
- Solution gap
- 20
- Willingness to pay
- 19
- Buildability
- 13
This idea targets a real, severe pain point in on-device AI with a clear gap for tiny, expressive, and easy-to-deploy models, and there are strong signals that developers would pay for a reliable solution. The technical complexity requires significant ML skill but leverages existing open-source work.
Value Equation
Dream outcome and how likely it feels, against the time and effort it costs.
Strong market demand for on-device TTS, excellent value proposition with tiny SOTA models, and a growing market. Differentiation is clear, but building and sustaining SOTA ML requires significant resources.
One-Person Business
Curiosity pull, identity fit, and a path from free value to paid for a solo creator.
A clear problem for a specific developer audience, but sustaining SOTA ML development and managing the open-source project for monetization will be complex for a solo builder.
Viral Frameworks
Hook strength, shareability, and how cheaply it can be tested.
Strong, measurable value prop for a specific developer audience, but needs more structured validation and careful business model execution around the open-source core.
Builder Lens
Evidence the problem exists, timing, defensibility, and a model that fits on a napkin.
Strong demand for solving real pain points with a specific target and narrow wedge, poised for increasing relevance in the future of AI.
Why this verdict
Five lenses, one composite. How scoring works
The angle
This weekend
- Focus on developer experience: make installation and integration exceptionally easy, especially compared to Coqui TTS.
- Actively engage with developers on GitHub, Hacker News, and relevant forums to gather feedback and build a community around the new models.
- Explore and validate potential commercialization paths (e.g., early enterprise partnerships, commercial licenses) alongside the open-source release.
- Document performance benchmarks rigorously, particularly against competitors in the 'tiny model, high expressivity, no GPU' niche.
Who is already there, emerging market
- Coqui TTS (XTTS-v2)Visit
An open-source deep learning toolkit for Text-to-Speech synthesis with voice cloning capabilities and multilingual support.
Pricing: Free (open-source), but commercial use restricted without specific licensing. Cloud inference costs may apply for API usage.
The world's first on-device, super-realistic TTS model with instant voice cloning, built on a compact 0.5B-parameter LLM backbone.
Pricing: Not explicitly stated, but marketed for on-device deployment suggesting potential one-time licensing or enterprise solutions.
A high-performance, open-source TTS model family for low-latency, production-grade voice applications, using a streamlined 350M-parameter architecture.
Pricing: Pricing for Resemble AI's commercial offerings starts at $0.01/second for standard voices and $0.015/second for cloned voices.
- Kokoro (indie-developed)Visit
A lightweight yet high-quality TTS model with just 82 million parameters, offering efficient running on modest hardware.
Pricing: Free (open-source, Apache 2.0 license).
A Llama-based TTS model available in multiple parameter versions (3B, 1B, 400M, 150M) optimized for natural, human-like speech.
Pricing: Not explicitly stated, likely enterprise-focused.
An ultra-fast, multilingual text-to-speech (TTS) model designed for real-time voice synthesis with sub-100ms latency and ultra-low VRAM usage.
Pricing: Likely API-based with free tier and enterprise options, specific numbers not publicly available for Lightning itself but Smallest.ai offers a TTS platform.
- Piper TTSVisit
A fast, local neural text-to-speech system that emphasizes small model size for on-device applications.
Pricing: Free (open-source).
- eSpeak-ngVisit
A compact open-source speech synthesizer that produces formant-synthesized speech.
Pricing: Free (open-source).
- Mozilla TTSVisit
An open-source deep learning-based speech synthesis engine designed for high-quality and customizable speech output.
Pricing: Free (open-source).
What they charge
Recent news
On-Device AI shifts more processing onto devices across IoT systems
IoT News, March 18 2026
KAIST-created SoulMate to live on your device as personal AI
Korea JoongAng Daily, March 17 2026
Exploring open-source tools for integrating text to speech in conversational AI
ElevenLabs, March 14 2026
Kitten TTS: Open-Source Voice Synthesis for Edge Devices
Sesame Disk, March 19 2026
The best text-to-speech software in 2026 - Product Hunt
Product Hunt, March 21 2026
Market signals
The market for on-device and small footprint text-to-speech (TTS) models is a rapidly growing niche within the larger AI industry, driven by the increasing demand for privacy, reduced latency, and lower operational costs in edge AI and IoT applications. Recent signals from chipmakers and smartphone manufacturers indicate a strong trend towards shifting more AI processing onto devices. This includes significant advancements in Neural Processing Units (NPUs) embedded in consumer chips. This shift is also influenced by the high operational costs associated with cloud AI at scale.
What frustrates people
- Coqui TTS has brutal installation challenges and a steep learning curve.
- Coqui TTS's original company shutdown created chaos that still affects usability.
- Traditional TTS models can sound mechanical and sluggish.
- Cloud TTS incurs unpredictable per-request billing and round-trip latency that kills conversational flow.
- The privacy concern of where user text data ends up with cloud TTS.
Semble – Code search for agents that uses 98% fewer tokens than grep
Hey HN! We (Stephan and Thomas) recently open-sourced Semble. We kept running into the same problem while using Claude Code on large codebases: when the agent can't find something directly, it falls back to grep, reading full files or launching subagents. This uses a lot of tokens, and often still misses the relevant code. There are existing tools for this, but they were either too slow to index on demand, needed API keys, or had poor retrieval quality.Semble is our solution for this. It combines static Model2Vec embeddings (using our latest static model: potion-code-16M) with BM25, fused via RRF and reranked with code-aware signals. Everything runs on CPU since there's no transformers involved. On our benchmark of ~1250 query/document pairs across 63 repos and 19 languages, it uses 98% fewer tokens than grep+read and reaches 99% of the retrieval quality of a 137M-parameter code-trained transformer, while being ~200x faster.Main features:- Token-efficient: 98% fewer tokens than grep+read- Fast: ~250ms to index a typical repo on our benchmark, ~1.5ms per query on CPU (very large repos may take longer)- Accurate: 0.854 NDCG@10, 99% of the best transformer setup we tested- MCP server: drop-in for Claude Code, Cursor, Codex, OpenCode- Zero config: no API keys, no GPU, no external servicesInstall in Claude Code with: claude mcp add semble -s user -- uvx --from "semble[mcp]" sembleOr check our README for other installation instructions, benchmarks, and methodology:Semble: https://github.com/MinishLab/sembleBenchmarks: https://github.com/MinishLab/semble/tree/main/benchmarksModel: https://huggingface.co/minishlab/potion-code-16MLet us know if you have any feedback or questions!
AI
Postgres extension for BM25 relevance-ranked full-text search
Last summer we faced a conundrum at my company, Tiger Data, a Postgres cloud vendor whose main business is in timeseries data. We were trying to grow our business towards emerging AI-centric workloads and wanted to provide a state-of-the-art hybrid search stack in Postgres. We'd already built pgvectorscale in house with the goal of scaling semantic search beyond pgvector's main memory limitations. We just needed a scalable ranked keyword search solution too.The problem: core Postgres doesn't provide this; the leading Postgres BM25 extension, ParadeDB, is guarded behind AGPL; developing our own extension appeared daunting. We'd need a small team of sharp engineers and 6-12 months, I figured. And we'd probably still fall short of the performance of a mature system like Parade/Tantivy.Or would we? I'd be experimenting long enough with AI-boosted development at that point to realize that with the latest tools (Claude Code + Opus) and an experienced hand (I've been working in database systems internals for 25 years now), the old time estimates pretty much go out the window.I told our CTO I thought I could solo the project in one quarter. This raised some eyebrows.It did take a little more time than that (two quarters), and we got some real help from the community (amazing!) after open-sourcing the pre-release. But I'm thrilled/exhausted today to share that pg_textsearch v1.0 is freely available via open source (Postgres license), on Tiger Data cloud, and hopefully soon, a hyperscalar near you:https://github.com/timescale/pg_textsearchIn the blog post accompanying the release, I overview the architecture and present benchmark results using MS-MARCO. To my surprise, we were not only able to meet Parade/Tantivy's query performance, but exceed it substantially, measuring a 4.7x advantage on query throughput at scale:https://www.tigerdata.com/blog/pg-textsearch-bm25-fu
AI
Consus (YC F24) – LLM gateway for companies handling sensitive data
Hey Hacker News! I'm Eric, the founder of Consus. We route all AI requests to ITAR restricted and controlled unclassified information (CUI) compliant infrastructure. Consus itself satisfies ITAR restrictions, and we recently completed our FedRAMP equivalency audit against the 410 FedRAMP High controls. FedRAMP High is the U.S. government’s strictest security baseline for cloud services handling sensitive unclassified data. A breach of this data could cause severe or catastrophic harm. Achieving FedRAMP equivalency, or authorization, is required before processing CUI information.Demo: https://youtu.be/6MYxJe6P4oYAerospace and defense is experiencing rapid growth. Unfortunately, most companies are struggling to use AI because of necessary data protections & compliance risk. These companies can move even faster if they could use the AI tools that most other industries take for granted.I spent most of my career prior to Consus in aerospace & defense. I launched national security payloads at SpaceX, and was an early employee at Castelion. I’ve watched first hand how compliance red-tape makes it difficult for companies to adopt new technologies and trust vendors that touch their sensitive data. For example, I worked with a satellite manufacturer that doesn’t allow any AI tools, even for non-sensitive work.Even as AI has become extremely useful over the past 9-12 months, I was surprised by how many aerospace & defense companies refused to touch it. They could see how it could be helpful, but they were scared about model providers training on their data, violating an ITAR rule, or risk losing new business because of a failed cyber assessment. These are all rational fears.There were a lot of ways to solve some/all of these problems, but we focused on building a compliant AI inference layer that can "plug" into any AI tool or workflow. Just as OpenRouter integrates with most tools, Consus can connect to tools like ChatGPT or Clau
AI
Questions about this idea?
FlyBot reads the scoring and gives you a second opinion on “Three new Kitten TTS models – smallest less than 25MB”.