Todoist Ramble, a new AI-powered method of capturing tasks by merely speaking in the task management software used by more than 50 million users, is now generally available.
Ramble, which was created by a small, AI-powered team, helps individuals and teams get ideas out of their thoughts and into action more quickly than typing by converting natural, unstructured speech into clean, organized activities with specifics like projects, dates, deadlines, priority, and length.
Ramble is intended for fast-capture situations, like going between meetings, driving home, or balancing life and work, where thoughts flow more quickly than fingers.
Todoist creates an organized list when users tap the Ramble waveform symbol and speak normally. Ramble refreshes the tasks in real time and allows basic voice modifications during capture.
“The best capture method is the one you’ll actually use,”
said Amir Salihefendić, Doist Founder and CEO.
“People often think in unstructured fragments. Ramble captures those ideas as they happen and turns them into actionable tasks in Todoist with no special syntax or mental overhead required.”
Ramble makes it simple to record updates, client requests, and follow-ups in Todoist for Teams while switching between meetings.
Tasks can be added to both Personal and Team workspaces in a single session; Ramble updates shared lists without requiring typing by routing each item to the appropriate project, assignee, and date based on what is spoken.
Ramble is one of the first productivity-related production deployments that employ Google’s Gemini 2.5 Flash Live model via Vertex AI for real-time speech recognition.
It is based on Todoist Assist. As users talk, audio is sent to Doist’s backend, where the model transcribes the voice and the system generates tasks by parsing entities and intent (such as projects, assignees, dates, repeating rules, priorities, and durations).
To add, change, or remove tasks in real time, the model makes structured tool calls. The service has minimal latency on desktop, iOS, and Android.
