Llamatik
A Kotlin Multiplatform library for running large language models, speech-to-text and image generation fully on-device across Android, iOS, desktop and WASM, built on llama.cpp, whisper.cpp and stable-diffusion.cpp.
Llamatik exposes llama.cpp for LLM text generation, whisper.cpp for speech-to-text and stable-diffusion.cpp for image generation behind one unified Kotlin API, so an app written once in Kotlin Multiplatform can run all three fully offline on Android, iOS, desktop, JVM or WASM, with no Python and no required server.
Its LLM support covers streaming and non-streaming generation, schema-constrained JSON output, embeddings, configurable context length and Flash Attention, KV-cache session save and resume, concurrent inference sessions, chat-template introspection, and multi-token speculative prediction on supported models. The speech component adds segment-aware transcription with timestamps, optional translation to English and speaker-turn detection, and the image component covers text-to-image and image-to-image generation. MIT licensed.