NEW DELHI — Users in India reported widespread problems with ChatGPT's speech-to-text feature Thursday, with voice input failing to process audio prompts and triggering error messages, prompting speculation about whether the disruption was limited to the country or part of a broader OpenAI service issue.

ChatGPT

The complaints surfaced prominently on social media, including a detailed post from tech user Tushar Dahiya on X that quickly drew attention. Dahiya shared a screenshot showing the feature not responding and wrote, "I think chatgpt for speech to text cdn is down in india, tried using their speech to text for giving inputs to chat, but it's not working." He followed up asking if the outage was India-specific or global.

As of mid-afternoon local time, monitoring sites like Downdetector showed no massive global spike in overall ChatGPT reports, but scattered user comments from India highlighted voice mode and audio transcription problems. Many described being unable to dictate messages in the mobile app or web interface, forcing them to type prompts manually. The issue appeared concentrated on real-time speech-to-text conversion rather than core chat functionality.

OpenAI had not issued an official statement on its status page or social channels by early evening, unlike in previous outages when the company quickly acknowledged problems. The lack of immediate confirmation left users guessing whether the disruption stemmed from a content delivery network (CDN) issue in India, a regional server problem or something more widespread affecting the voice features rolled out in recent updates.

ChatGPT's speech-to-text and voice conversation capabilities, powered by OpenAI's Whisper model and integrated with advanced audio processing, have become popular for hands-free use, accessibility and quick idea capture. The feature allows users to speak naturally and receive transcribed text or engage in spoken dialogues. India, with its large English-speaking tech and student population, ranks among the fastest-growing markets for the AI chatbot, making any hiccup in voice input particularly noticeable.

This is not the first time ChatGPT has faced regional or feature-specific glitches. Earlier in 2026, OpenAI dealt with brief outages in February that affected conversation loading and login across the U.S. and parts of Asia, though those were quickly resolved within hours. In late 2025, a Cloudflare-related disruption knocked out access globally, including in India. Thursday's reports, however, seemed narrower — focused on speech-to-text rather than the entire platform.

Tech observers noted that CDN problems are common culprits for geographically isolated issues. OpenAI relies on multiple global partners for low-latency delivery of AI models, including audio processing. A temporary overload, maintenance or routing error in Indian data centers could explain why users elsewhere reported normal service. Some Indian users said switching to a VPN temporarily bypassed the problem, suggesting a localized routing or edge-server issue.

The timing added frustration for many. April marks the start of the academic year in parts of India, with students and professionals relying on ChatGPT for study aids, research and productivity tools. Voice input is especially valued during commutes or multitasking. "It's annoying when you're driving or cooking and suddenly can't speak to it," one Mumbai-based software engineer posted on X alongside similar complaints.

OpenAI has expanded voice features significantly since launching them in 2024, integrating more natural-sounding responses and multilingual support. The company touts the technology as one of its most advanced, but it also introduces dependencies on real-time audio pipelines that can be sensitive to network conditions. In India, where mobile data and Wi-Fi quality vary widely, such features are both highly useful and vulnerable to hiccups.

No widespread reports emerged from the United States, Europe or other major markets, reinforcing the possibility of a regional CDN or India-specific configuration problem. OpenAI's global status page, which tracks incidents for ChatGPT, the API and related services, showed no active alerts for April 2 as of the latest checks. The company's history shows it typically posts updates within minutes of detecting elevated errors.

For affected users, workarounds included typing prompts directly, using the desktop version (which some said worked better) or waiting a few hours for potential resolution. OpenAI has not offered credits or apologies for minor glitches in the past, but prolonged issues sometimes prompt compensation for Plus and Team subscribers.

The episode highlights growing dependence on AI tools for everyday tasks. Millions in India use ChatGPT for everything from coding assistance to language practice and content creation. Any disruption, even brief, ripples through classrooms, offices and freelance workflows. Analysts estimate OpenAI's user base in India has grown exponentially, driven by free-tier access and integration with WhatsApp and other local apps.

Broader context includes OpenAI's ongoing efforts to scale infrastructure amid exploding demand. The company has invested heavily in data centers and partnerships with cloud providers to handle voice, image and video features. Still, occasional outages underscore the challenges of running real-time AI at planetary scale.

Thursday's reports come amid a relatively stable period for ChatGPT after several high-profile incidents earlier in the year. In February, thousands complained of "conversation not loading" errors, prompting OpenAI to confirm and fix the problem within hours. Those outages affected core text chats more than voice features. India-specific complaints have surfaced before, often tied to high traffic during peak evening hours.

Experts advise users experiencing voice issues to check their internet connection, update the app, clear cache or try incognito mode. Restarting the device sometimes resolves temporary glitches. For persistent problems, reporting through the in-app feedback or OpenAI's help center helps the company identify patterns.

As the day progressed, some users noted the feature starting to work intermittently, suggesting the issue might be resolving on its own or through backend adjustments. Others continued posting screenshots of error messages, keeping the topic alive on Indian tech forums and X.

OpenAI, valued at over $150 billion and backed by Microsoft, has transformed how people interact with AI. Its voice mode, once a premium feature, is now central to the product. Disruptions like Thursday's remind users — and investors — of the fragility of even the most advanced cloud services.

For now, the speech-to-text outage appears confined to India and limited to audio input, with no impact on text-based chats or other OpenAI tools like DALL-E or Sora. The company's silence suggests engineers are investigating without classifying it as a major incident.

Users in India are urged to monitor OpenAI's official status page or Downdetector for updates. In the meantime, alternative AI apps with voice features, such as Google's Gemini or Microsoft's Copilot, offered temporary substitutes for some.

The incident, though minor, underscores India's rising importance in the global AI landscape. With one of the world's largest digital populations, any service glitch here draws quick attention and tests OpenAI's ability to deliver consistent performance across diverse networks.

As evening approached in India, many hoped for a swift fix so they could resume hands-free interactions with ChatGPT. Whether the problem proves regional or part of a stealthier global tweak remains unclear pending an official explanation.

OpenAI has built a reputation for rapid recovery from outages, often restoring service before widespread panic sets in. Thursday's event followed that pattern so far, with no confirmation of a full-blown disruption but enough user frustration to spark online discussion.

For a tool that millions turn to daily, even short-lived speech-to-text hiccups serve as a reminder of how intertwined modern work and learning have become with AI infrastructure — and how quickly a single feature can affect routines when it falters.