Voice and Brush: How AI Is Changing the Way Painters Sketch and Think

An Old Habit Gets a New Voice

For years, voice input meant holding a button, speaking, and then fixing the mistakes. Painters used it to jot down notes about color mixtures or composition ideas while their hands were covered in paint. It worked, barely. The errors, the missing commas, the awkward phrasing—it all had to be cleaned up later.

Then large language models showed up. Suddenly, the computer could understand not just the words but the intent. It could strip out the ums and ahs, fix the grammar, and even adapt the tone to fit the context. That changed everything.

In China, this shift has been especially visible. Doubao's keyboard made voice the core feature, handling dialects, mixed Chinese-English speech, and low-bandwidth situations. Qwen added voice input to its PC app in May, going beyond simple transcription to clean up run-on sentences and structure the output. WeChat's keyboard followed, offering structured text from spoken ideas.

The result? What used to be a rough draft you had to edit now becomes a message you can actually send. For a painter, that same flow applies to studio notes, client emails, or even grant applications. You talk, the machine listens, and the text comes out ready.

Voice Cursor: A New Name in the Game

Into this crowded field steps Voice Cursor, a startup that just raised $8 million in seed funding, led personally by Su Hua, the founder of Kuaishou. The team is small, but the credentials are strong.

The founder, Chen Long, has a background that reads like a tech résumé from central casting. He worked on NLP at Baidu, spent time at Microsoft and Square, then founded a company called Avocado Tech, which built AI interview software and attracted investment from Sequoia Capital China, GSR Ventures, and others. After that company was acquired and folded into ByteDance, he rose to become vice president of product for Feishu, the enterprise collaboration tool.

The other co-founder, Henry Song, studied computer science at Berkeley and has been tinkering with AI projects since his student days. He had a hand in GPTBang and founded Skylow, which experimented with using AI to transform video learning. Berkeley's official channels even mentioned his high school software that caught the attention of a Fortune 500 company.

Voice Cursor works like any other voice input tool at first. You press a hotkey and start talking. You don't need to have the sentence fully formed in your head. You can repeat yourself, pause, change your mind mid-sentence—the system cleans it all up and produces text that reads like a direct message, not a transcript.

That part isn't unique. What sets Voice Cursor apart is what happens after the text appears. Select a block of text and say “make it shorter.” The system rewrites it. Say “make it friendlier,” and it adjusts the tone. The name “Cursor” points to this idea: the voice follows your cursor. Wherever you're working—an email, a document, a chat window—the voice is right there, ready to edit and refine.

Context Is Everything

The team emphasizes context. The same sentence sounds different in a studio notebook versus a formal proposal. If you've already mentioned a specific pigment or a client's name earlier in the text, the system can use that to understand what you're saying now. That's not a moat, but it's a start.

The harder part is figuring out intent. When you select a sentence and keep talking, is that a follow-up thought to append, or do you want to replace what's there? Voice Cursor tries to read that from the situation.

For painters, this matters more than you might think. A lot of creative work involves talking to yourself—muttering about a composition, considering whether to push the blue further, deciding if the brushwork is too tight. Voice tools can capture that thinking in real time, turn it into clear notes, and help you organize it into a plan for the day.

Hardware: The VoiceKit Stick

Voice Cursor also introduced a small hardware companion called VoiceKit. It's a pocket-sized device that pairs with the software, letting you speak into it to input text, edit, click, and send commands. It's not a standalone AI gadget; it depends on the Voice Cursor app. The underlying hardware is based on the M5Stick S3, and if you already own one of those, you can flash the open firmware and use it.

The pricing is straightforward: $144 a year for the software, which includes a free VoiceKit Stick. Or you can buy the stick separately for $99, but you still need the subscription.

Early numbers are modest—100 users in the first week, and all of them came back the next day. For a product still testing its habits, that retention is the metric the team watches closest. It's a tiny signal, but it suggests the stick isn't just a gimmick; people are actually using it.

Why Voice Works for Artists

Chen Long has talked about his own workflow on social media. He records ideas with voice as soon as they pop into his head, hands them to Claude for further thinking, and then executes at the cursor. For him, the most fragile moment is when an idea moves from his brain to the keyboard. He says the mind holds a rich mix of background, details, and nuance, but typing flattens it. You compress your thoughts because typing is slow.

That's true for painters too. When you're standing at the easel, brush in hand, the last thing you want to do is wash up, find your phone, and type a note. But if you can just talk—say, “the light on the left edge is too yellow, try a cooler tone, maybe mix a touch of cerulean”—the machine can capture the full thought. Later, you can clean it up or send it to an AI assistant for further exploration.

Voice also offers a way to explain more, faster. With typing, you tend to strip an idea down to its bare bones. You lose the why, the constraints, the examples. With voice, you can give the model more to work with. A prompt like “make the background darker” becomes “the background feels too bright, I want it to recede, maybe deepen the shadows and reduce the contrast so the figure stands out more.” The model gets more context, and the result is closer to what you actually wanted.

The Competition Is Already Crowded

Voice Cursor isn't entering an empty room. Typeless is the most direct comparison—it also cleans up speech, handles mid-sentence changes, and adjusts tone per app. Wispr Flow has been at this longer, with more funding and a larger user base. It uses dictionaries and surrounding text to improve recognition of technical terms and proper nouns, and you can set different writing styles for different apps.

So Voice Cursor can't claim to have invented a new category. But it's betting that the real differentiation will come from how the system handles what you say, not just how accurately it transcribes. As base models improve, everyone will get good at transcription. The edge will come from understanding intent and managing the messy, in-the-moment nature of human speech.

There's also the ecosystem question. Chinese tech giants have built-in distribution through their keyboards and apps. Independent startups like Voice Cursor have to win on experience and cross-platform polish. That's a harder road, but it forces focus.

A New Signal for How We Work

For painters, the rise of voice AI is more than a convenience. It's a new way to keep a sketchbook of ideas without interrupting the flow. You can narrate your process, capture a thought about a technique, or dictate a quick list of supplies while your hands are busy. The machine cleans it up, organizes it, and hands it back to you ready to use.

The bigger picture is that AI is getting better at doing things, which means the bottleneck is shifting to how well we explain what we want. Voice is one way to communicate more of our intent in less time. It's not about replacing the keyboard entirely—some people like the rhythm of typing. But for those moments when your hands are covered in paint or your mind is racing faster than your fingers, voice is becoming a real option.

Voice Cursor is still early, and the competition is fierce. But its existence is a signal. The way we tell machines what to do is becoming a product opportunity in itself. And for anyone who works with their hands—painters, sculptors, ceramicists—that's worth paying attention to.