CELESTOS
CELESTOS
Everything you've worked on, researched, thought through, and captured — synthesised into a two-host conversation and rendered as audio on your machine. Put it on while you commute. Play it back before you return to a project. Your memory, always with you, even when you're not at a screen.
Celestos doesn't produce a robot reading your notes aloud. It synthesises your sessions, research, and memory into a two-host dialogue — one explains, one questions — so what you hear sounds like a real discussion about the work you've been doing. It makes your own knowledge easier to absorb than reading it.
Pick a project, a single session, a topic, or a bundle of files. Celestos reads everything you've fed it on that subject — notes, conversations, documents, voice memos, research.
The local LLM synthesises the content into a two-host dialogue. Host A explains and recaps. Host B asks the questions a listener would ask. Timed to your chosen length.
The embedded ONNX TTS engine renders each speaker's lines with distinct voices. No cloud audio service. No API key. WAV output generated entirely on your machine.
The file goes to your output folder. Listen in any audio player, AirDrop it to your phone, or queue it up for your commute. Regenerate whenever your project grows.
Most AI tools lose you the moment you close the window. With Memory Podcasts, your project context travels with you as audio. Listen before a meeting to remember where you left off. Listen during a run to absorb a research thread. Your Celestos knowledge doesn't stay on your desk — it goes wherever you go.
Generate a short catch-up from yesterday's sessions on your phone during your commute. By the time you open Celestos, you already know where you were.
Ask for a 3-minute brief on a client project before you walk into the call. Celestos pulls from everything it knows about them and narrates it back to you.
Set a Spark to auto-generate a Friday podcast from the week's AI sessions. Listen on the weekend. Stay on top of everything across all your projects without sitting down to review it.
Generate a podcast from a project and send the WAV to a collaborator. They get your full context — the research, the decisions, the reasoning — in a format they can listen to on the go.
Every part of the pipeline — script generation, voice synthesis, audio rendering — runs on your machine using local models and the embedded ONNX TTS engine. Nothing leaves your device. No cloud API call. No third-party service processing your private sessions. Your knowledge stays yours.
The dialogue script is written by the same local model powering your chat. No external API call.
Speech is synthesised by a built-in neural TTS engine running on-device. No ElevenLabs, no Google TTS, no cloud.
The finished audio is a plain WAV file on your machine. Play it anywhere. Keep it forever. Delete it whenever you want.