Voice and vision are the senses of the 2026 cohort: 27 startups teaching AI to hear, speak and see. 17 work on voice and audio, 10 on vision. They are small categories next to agents (137) and vertical AI (149), but they supply the interfaces everyone else builds on. This page lists all 27.

1. All 27 companies

Newest batch first, category on every row. Each links to a company page with founders and the full story.

CompanyWhat it doesCategoryBatchHQ
AgeospatialPhysical intelligence for every decision with Geospatial AI.AI VisionYC Fall 2026San Francisco, CA, USA; Remote
antimattrvoice-first hardware interface for human-agent interactionAI Voice and AudioYC Fall 2026San Francisco, CA, USA
AudoraAudiobooks where every character has their own voice.AI Voice and AudioYC Fall 2026San Francisco, CA, USA
BayworksIntelligence that keeps America’s vehicles runningAI Voice and AudioYC Fall 2026San Francisco, CA, USA
DeepMarkA live watermark inside AI voice that survives phone callsAI Voice and AudioYC Fall 2026San Francisco, CA, USA
PraxisAI agents that run law firm operations.AI Voice and AudioYC Fall 2026San Francisco, CA, USA
Sona8Voice agents survey employees to map, improve & automate workflowsAI Voice and AudioYC Fall 2026San Francisco, CA, USA
The Agentic Data Co.Training Data For Speech Models.AI Voice and AudioYC Fall 2026San Francisco, CA, USA
VerdunThe world's most advanced security monitoring serviceAI VisionYC Fall 2026San Francisco, CA, USA
Dialogusvoice AI agents that learn how your company actually worksAI Voice and AudioYC Summer 2026San Francisco, CA, USA
DreamUsing VLMs to Catch DamageAI VisionYC Summer 2026San Francisco, CA, USA
EliteHealthcare AIAI Voice and AudioYC Summer 2026San Francisco, CA, USA
EvergroveInstant healthcare for all, starting with work compAI Voice and AudioYC Summer 2026New York City, NY, USA
GrocaloAI brain for creators that runs their social contentAI VisionYC Summer 2026Los Angeles, CA, USA
OperonAgentic data layer for manufacturing & process industries.AI VisionYC Summer 2026San Francisco, CA, USA
PlutoLinkedIn for agentsAI Voice and AudioYC Summer 2026San Francisco, CA, USA
SidekickAI agent that manages manufacturing operations over textAI VisionYC Summer 2026San Francisco, CA, USA
SpekoOpenRouter for Voice AI: one API, every call on the best speech modelAI Voice and AudioYC Summer 2026San Francisco, CA, USA
SubVysionAutonomous rovers to make the ‘Google Maps’ of underground utilitiesAI VisionYC Summer 2026San Francisco, CA, USA
Callab AIHuman-centered AI for contact centers.AI Voice and AudioYC Spring 2026San Francisco, CA, USA
Keyframe LabsBring AI to life with the world's best interactive avatarsAI Voice and AudioYC Spring 2026Los Angeles, CA, USA
Miso LabsEmotive foundation voice modelsAI Voice and AudioYC Spring 2026San Francisco, CA, USA
BidflowAI Takeoffs for ElectricalAI VisionYC Winter 2026New York City, NY, USA
KitaUnderwrite borrowers around the world in minutesAI VisionYC Winter 2026San Francisco, CA, USA
Piris LabsThe Networking Layer of AIAI VisionYC Winter 2026San Francisco, CA, USA
PocketTake Notes in the Real WorldAI Voice and AudioYC Winter 2026San Francisco, CA, USA
Talking ComputersReal-time interaction models for voice agentsAI Voice and AudioYC Winter 2026San Francisco, CA, USA

2. Voice leans late, vision leans summer

Voice and audio builds through the year: 2 in Winter, 3 in Spring, 5 in Summer and 7 in Fall 2026, as realtime voice models got cheaper to run. Vision is flatter: 3 in Winter, none in Spring, 5 in Summer and 2 in Fall. The vision companies skew industrial (inspection, documents, physical monitoring), while voice skews to agents that talk: support calls, sales calls and phone workflows.

3. Why two small categories get a page

Because interfaces compound. A support agent startup needs a voice; a warehouse robot needs vision. When these 27 raise their game, the categories built on them move too. Both lists stay current on the voice and audio and vision category pages.

4. Where the 27 plug in

Voice plugs into conversations: phone support, sales calls, scheduling, interviews, any workflow that still runs on spoken words between people. Vision plugs into the physical and the documentary: inspecting parts and shelves, reading forms and drawings, watching spaces. Notice how many of the 27 sell to a named workflow rather than a technology. A voice agent for dental appointment reminders and a vision system for electrical drawings are not demos looking for a problem; they are narrow, priced and already aimed.

A last observation from the table: almost none of the 27 describe themselves as model companies. They sell calls answered, parts inspected, forms read and shelves counted. The models are inside, unnamed. That is usually the sign of a category leaving the demo phase: the technology stops being the headline and becomes the plumbing.

Source: AIToolGuide 2026 startups database (513 AI companies), compiled from Y Combinator 2026 batch listings and the TrustMRR and Indie Hackers marketplaces.