# an AI transcription service

**AI transcription is crowded with active open-source tools, APIs, and complaints about accuracy, latency, and missing features.**

Confidence: high  ·  116 signals  ·  checked 2026-08-01

[Full report with live links](https://showsitsworks.com/idea/an-ai-transcription-service)

## Who is already building this

*From data*

- Meetily is a privacy-first, local AI meeting assistant with 27,920 stars and a last push on 2026-06-05. It does live transcription, speaker diarization, and summarization.
  - [Zackriya-Solutions/meetily](https://github.com/Zackriya-Solutions/meetily) `github` — 2024-12-26
- Omi has 13,100 stars and was pushed on 2026-08-02. It listens to conversations and tells you what to do, so transcription is part of a broader assistant layer.
  - [BasedHardware/omi](https://github.com/BasedHardware/omi) `github` — 2024-03-22
- Vexa has 2,634 stars and was pushed on 2026-07-30. It is an open-source meeting transcription API for Google Meet, Microsoft Teams, and Zoom, with real-time WebSocket transcripts and self-hosting.
  - [Vexa-ai/vexa](https://github.com/Vexa-ai/vexa) `github` — 2025-02-07
- Whishper has 3,049 stars and was pushed on 2026-07-31. It transcribes, translates, and edits subtitles 100% locally with a web UI.
  - [pluja/whishper](https://github.com/pluja/whishper) `github` — 2023-08-26
- Scriberr has 2,875 stars and was pushed on 2026-06-01. It is a self-hosted AI audio transcription project.
  - [rishikanthc/Scriberr](https://github.com/rishikanthc/Scriberr) `github` — 2024-10-04
- TypeWhisper has 1,660 stars and was pushed on 2026-08-02. It offers local speech-to-text for macOS, with optional cloud support.
  - [TypeWhisper/typewhisper-mac](https://github.com/TypeWhisper/typewhisper-mac) `github` — 2026-02-12
- The npm package @ai-sdk/elevenlabs had 740,566 downloads last month and includes batch and realtime transcription support. The npm package @ai-sdk/deepgram had 669,249 downloads last month and also exposes transcription model support.
  - [npm: @ai-sdk/elevenlabs](https://www.npmjs.com/package/@ai-sdk/elevenlabs) `registries` — 2026-07-31
  - [npm: @ai-sdk/deepgram](https://www.npmjs.com/package/@ai-sdk/deepgram) `registries` — 2026-07-31

## What people actually say

*From data*

- "I was tired of waiting hours for transcription services and didn't want to upload sensitive recordings to the cloud." A Hacker News founder says the pain is speed plus privacy.
  - [Show HN: Scriber Pro – Offline AI transcription for macOS](https://news.ycombinator.com/item?id=45591222) `hackernews` — 2025-10-15
- "Most notably, transcription is sl..." A Hacker News builder of voice chat with AI says the cascaded transcription step is a major limitation.
  - [Show HN: Real-time voice chat with AI, no transcription](https://news.ycombinator.com/item?id=39759147) `hackernews` — 2024-03-19
- "I don't want every dictation to automatically go through AI." A GitHub issue asks for raw transcription without forced post-processing.
  - [Feature Request: Separate Hotkeys for Raw Transcription and AI Processing + Screen Context Integration](https://github.com/EpicenterHQ/epicenter/issues/567) `githubissues` — 2025-07-11
- "I'd spent quite a bit of time looking." An HN commenter says they searched for a 24/7 real-time voice transcription tool and still had not found a good fit.
  - [Ask HN: Looking for a 24-7 Real-Time Voice Transcription Tool](https://news.ycombinator.com/item?id=37544667) `intent` — 2023-09-17
- "transcription session is successfully created, but I am not receiving the transcription text." A Discourse user is hitting a live transcription delivery failure.
  - [Realtime transcription issue](https://community.openai.com/t/realtime-transcription-issue/1150994) `discourse` — 2025-03-24

## People asking to be sold to

*From data*

- "Looking for a 24-7 Real-Time Voice Transcription Tool" is a direct ask HN intent signal from someone describing an unmet need.
  - [Ask HN: Looking for a 24-7 Real-Time Voice Transcription Tool](https://news.ycombinator.com/item?id=37544667) `intent` — 2023-09-17
- "What is your recommended speech to text/audio transcription tool?" is another intent signal asking for a tool choice rather than a general discussion.
  - [Ask HN: What is your recommended speech to text/audio transcription tool?](https://news.ycombinator.com/item?id=36302434) `intent` — 2023-06-12
- "The current approaches all do a cascaded approach... This approach is easy to get started with, but requires lots of complexity" shows a willingness to replace the current stack with something simpler.
  - [Show HN: Real-time voice chat with AI, no transcription](https://news.ycombinator.com/item?id=39759147) `hackernews` — 2024-03-19
- "I'd wish there was a better faster one built on top of Whisper" is a stated want for a better product than current options.
  - [Ask HN: What is your recommended speech to text/audio transcription tool?](https://news.ycombinator.com/item?id=36302434) `intent` — 2023-06-12

## Where the opening is

*Model estimate*

- The clearest gap is not basic speech-to-text. It is a dependable 24/7, low-latency transcription workflow that handles long sessions, live delivery, and optional raw-vs-AI post-processing without breaking.
  - [Ask HN: Looking for a 24-7 Real-Time Voice Transcription Tool](https://news.ycombinator.com/item?id=37544667) `intent` — 2023-09-17
  - [Realtime transcription issue](https://community.openai.com/t/realtime-transcription-issue/1150994) `discourse` — 2025-03-24
  - [Feature Request: Separate Hotkeys for Raw Transcription and AI Processing + Screen Context Integration](https://github.com/EpicenterHQ/epicenter/issues/567) `githubissues` — 2025-07-11
- Several tools are local-first or self-hosted, but the signals do not show a dominant product focused on always-on transcription for people who need long-running capture, editable output, and strong privacy in one workflow.
  - [Zackriya-Solutions/meetily](https://github.com/Zackriya-Solutions/meetily) `github` — 2024-12-26
  - [pluja/whishper](https://github.com/pluja/whishper) `github` — 2023-08-26
  - [rishikanthc/Scriberr](https://github.com/rishikanthc/Scriberr) `github` — 2024-10-04
  - [TypeWhisper/typewhisper-mac](https://github.com/TypeWhisper/typewhisper-mac) `github` — 2026-02-12
- The data also suggests a feature gap around reliability controls: word timestamps, response format options, chunking, multilingual output, and failure handling keep coming up in issue threads.
  - [@ai-sdk/openai transcription response_format is missing from the providerOptions](https://github.com/vercel/ai/issues/6562) `githubissues` — 2025-05-31
  - [@ai-sdk/openai transcription providerOptions timestamp_granularities is broken](https://github.com/vercel/ai/issues/6561) `githubissues` — 2025-05-31
  - [Audio Transcription API chunking_strategy option](https://community.openai.com/t/audio-transcription-api-chunking-strategy-option/1260804) `discourse` — 2025-05-14
  - [Enable other languages for meeting AI transcription](https://github.com/calcom/cal.diy/issues/21727) `githubissues` — 2025-06-06
  - [Recurring `TypeError: 'tuple' object has no attribute 'lower'` from `gpt-4o` / `gpt-4o-mini` transcription API (Streaming)](https://community.openai.com/t/recurring-typeerror-tuple-object-has-no-attribute-lower-from-gpt-4o-gpt-4o-mini-transcription-api-streaming/1321082) `discourse` — 2025-07-21
  - [Why does transcription_session.update cause an error? Realtime transcription API bug](https://community.openai.com/t/why-does-transcription-session-update-cause-an-error-realtime-transcription-api-bug/1370560) `discourse` — 2025-12-31

## How big the market might be

*Model estimate*

- This search found 12 named open-source projects, and 11 of them are still active. One is slow, but none look dead from the data given.
  - [Zackriya-Solutions/meetily](https://github.com/Zackriya-Solutions/meetily) `github` — 2024-12-26
  - [BasedHardware/omi](https://github.com/BasedHardware/omi) `github` — 2024-03-22
  - [tmoroney/auto-subs](https://github.com/tmoroney/auto-subs) `github` — 2023-03-15
  - [pluja/whishper](https://github.com/pluja/whishper) `github` — 2023-08-26
  - [rishikanthc/Scriberr](https://github.com/rishikanthc/Scriberr) `github` — 2024-10-04
  - [Vexa-ai/vexa](https://github.com/Vexa-ai/vexa) `github` — 2025-02-07
  - [sindresorhus/awesome-whisper](https://github.com/sindresorhus/awesome-whisper) `github` — 2023-05-10
  - [floneum/kalosm](https://github.com/floneum/kalosm) `github` — 2023-05-24
  - [Natively-AI-assistant/natively-cluely-ai-assistant](https://github.com/Natively-AI-assistant/natively-cluely-ai-assistant) `github` — 2026-01-27
  - [kaixxx/noScribe](https://github.com/kaixxx/noScribe) `github` — 2023-05-12
  - [TypeWhisper/typewhisper-mac](https://github.com/TypeWhisper/typewhisper-mac) `github` — 2026-02-12
  - [royshil/obs-localvocal](https://github.com/royshil/obs-localvocal) `github` — 2023-08-10
  - [amicalhq/amical](https://github.com/amicalhq/amical) `github` — 2025-05-08
  - [silverstein/minutes](https://github.com/silverstein/minutes) `github` — 2026-03-18
  - [alesaccoia/VoiceStreamAI](https://github.com/alesaccoia/VoiceStreamAI) `github` — 2023-12-26
- There were 35 first-hand complaints, 4 unsolved Stack Exchange questions, and 2 stated wants. That is enough to show repeated friction, not a one-off issue.
  - [Open AI Whisper is returning the transcription in English instead of the native language](https://stackoverflow.com/questions/73929778/open-ai-whisper-is-returning-the-transcription-in-english-instead-of-the-native) `stackexchange` — 2022-10-02
  - [How to transcribe Whisper Open AI transcription for large video files larger than 25mb using python?](https://stackoverflow.com/questions/76417335/how-to-transcribe-whisper-open-ai-transcription-for-large-video-files-larger-tha) `stackexchange` — 2023-06-06
  - [Using Whisper AI for transcription from Polish - encoding error?](https://stackoverflow.com/questions/78478908/using-whisper-ai-for-transcription-from-polish-encoding-error) `stackexchange` — 2024-05-14
  - [Is there a way to disable content filteration in Azure OpenAi or to process the prompt before using it](https://stackoverflow.com/questions/79494887/is-there-a-way-to-disable-content-filteration-in-azure-openai-or-to-process-the) `stackexchange` — 2025-03-08
- The largest usage signals are very large: @ai-sdk/elevenlabs at 740,566 monthly downloads and @ai-sdk/deepgram at 669,249 monthly downloads.
  - [npm: @ai-sdk/elevenlabs](https://www.npmjs.com/package/@ai-sdk/elevenlabs) `registries` — 2026-07-31
  - [npm: @ai-sdk/deepgram](https://www.npmjs.com/package/@ai-sdk/deepgram) `registries` — 2026-07-31
- Matching discussion appeared across at least 5 community forums: community.home-assistant.io, community.openai.com, discuss.elastic.co, discuss.python.org, and meta.discourse.org.
  - [GPT-Live-Transcribe and GPT-Transcribe: Two New Transcription Models in the API](https://community.openai.com/t/gpt-live-transcribe-and-gpt-transcribe-two-new-transcription-models-in-the-api/1388318) `discourse` — 2026-07-29
  - [Realtime transcription issue](https://community.openai.com/t/realtime-transcription-issue/1150994) `discourse` — 2025-03-24
  - [Audio Transcription API chunking_strategy option](https://community.openai.com/t/audio-transcription-api-chunking-strategy-option/1260804) `discourse` — 2025-05-14
  - [Recurring `TypeError: 'tuple' object has no attribute 'lower'` from `gpt-4o` / `gpt-4o-mini` transcription API (Streaming)](https://community.openai.com/t/recurring-typeerror-tuple-object-has-no-attribute-lower-from-gpt-4o-gpt-4o-mini-transcription-api-streaming/1321082) `discourse` — 2025-07-21
  - [Why does transcription_session.update cause an error? Realtime transcription API bug](https://community.openai.com/t/why-does-transcription-session-update-cause-an-error-realtime-transcription-api-bug/1370560) `discourse` — 2025-12-31
  - [Enable other languages for meeting AI transcription](https://github.com/calcom/cal.diy/issues/21727) `githubissues` — 2025-06-06

## What could go wrong

*Model estimate*

- This is a crowded market with strong incumbents and many adjacent open-source tools, so a generic transcription API will be hard to differentiate.
  - [Zackriya-Solutions/meetily](https://github.com/Zackriya-Solutions/meetily) `github` — 2024-12-26
  - [pluja/whishper](https://github.com/pluja/whishper) `github` — 2023-08-26
  - [rishikanthc/Scriberr](https://github.com/rishikanthc/Scriberr) `github` — 2024-10-04
  - [Vexa-ai/vexa](https://github.com/Vexa-ai/vexa) `github` — 2025-02-07
  - [TypeWhisper/typewhisper-mac](https://github.com/TypeWhisper/typewhisper-mac) `github` — 2026-02-12
  - [amicalhq/amical](https://github.com/amicalhq/amical) `github` — 2025-05-08
  - [silverstein/minutes](https://github.com/silverstein/minutes) `github` — 2026-03-18
- Accuracy and hallucination risk is a real trust problem, especially in regulated or high-stakes use cases.
  - [AI-powered transcription tool used in hospitals invents things no one ever said](https://news.ycombinator.com/item?id=41954130) `hackernews` — 2024-10-26
  - [Researchers say AI-powered transcription tool used in hospitals invents things](https://news.ycombinator.com/item?id=41953587) `hackernews` — 2024-10-26
  - [AI transcription tools 'hallucinate,' too](https://news.ycombinator.com/item?id=40176169) `hackernews` — 2024-04-27
  - [AI transcription tool 'hallucinates' medical interactions](https://news.ycombinator.com/item?id=42834898) `hackernews` — 2025-01-26
- Platform and provider friction is another risk: developers are already dealing with provider option bugs, chunking issues, language handling, and realtime API problems.
  - [Experimental transcription using Vercel AI Gatway](https://github.com/vercel/ai/issues/13504) `githubissues` — 2026-03-16
  - [@ai-sdk/openai transcription response_format is missing from the providerOptions](https://github.com/vercel/ai/issues/6562) `githubissues` — 2025-05-31
  - [@ai-sdk/openai transcription providerOptions timestamp_granularities is broken](https://github.com/vercel/ai/issues/6561) `githubissues` — 2025-05-31
  - [Realtime transcription issue](https://community.openai.com/t/realtime-transcription-issue/1150994) `discourse` — 2025-03-24
  - [Audio Transcription API chunking_strategy option](https://community.openai.com/t/audio-transcription-api-chunking-strategy-option/1260804) `discourse` — 2025-05-14
  - [Enable other languages for meeting AI transcription](https://github.com/calcom/cal.diy/issues/21727) `githubissues` — 2025-06-06
  - [Recurring `TypeError: 'tuple' object has no attribute 'lower'` from `gpt-4o` / `gpt-4o-mini` transcription API (Streaming)](https://community.openai.com/t/recurring-typeerror-tuple-object-has-no-attribute-lower-from-gpt-4o-gpt-4o-mini-transcription-api-streaming/1321082) `discourse` — 2025-07-21
  - [Why does transcription_session.update cause an error? Realtime transcription API bug](https://community.openai.com/t/why-does-transcription-session-update-cause-an-error-realtime-transcription-api-bug/1370560) `discourse` — 2025-12-31
- Some demand may already be met by local-first tools and by major API providers, which makes a standalone service vulnerable to pricing and model commoditization.
  - [npm: @ai-sdk/deepgram](https://www.npmjs.com/package/@ai-sdk/deepgram) `registries` — 2026-07-31
  - [npm: @ai-sdk/elevenlabs](https://www.npmjs.com/package/@ai-sdk/elevenlabs) `registries` — 2026-07-31
  - [Zackriya-Solutions/meetily](https://github.com/Zackriya-Solutions/meetily) `github` — 2024-12-26
  - [pluja/whishper](https://github.com/pluja/whishper) `github` — 2023-08-26
  - [TypeWhisper/typewhisper-mac](https://github.com/TypeWhisper/typewhisper-mac) `github` — 2026-02-12

## What to do this week

*Model estimate*

- Pick one narrow wedge: 24/7 real-time transcription for long-running calls or meetings, with privacy-first defaults and a hard focus on reliability.
  - [Ask HN: Looking for a 24-7 Real-Time Voice Transcription Tool](https://news.ycombinator.com/item?id=37544667) `intent` — 2023-09-17
  - [Show HN: Scriber Pro – Offline AI transcription for macOS](https://news.ycombinator.com/item?id=45591222) `hackernews` — 2025-10-15
  - [Realtime transcription issue](https://community.openai.com/t/realtime-transcription-issue/1150994) `discourse` — 2025-03-24
- Make raw transcription and AI post-processing separate toggles, because one user explicitly asked for that split.
  - [Feature Request: Separate Hotkeys for Raw Transcription and AI Processing + Screen Context Integration](https://github.com/EpicenterHQ/epicenter/issues/567) `githubissues` — 2025-07-11
- Support long-audio chunking, word-level timestamps, and multilingual output on day one, because those are recurring pain points in the issue and Stack Exchange signals.
  - [@ai-sdk/openai transcription response_format is missing from the providerOptions](https://github.com/vercel/ai/issues/6562) `githubissues` — 2025-05-31
  - [@ai-sdk/openai transcription providerOptions timestamp_granularities is broken](https://github.com/vercel/ai/issues/6561) `githubissues` — 2025-05-31
  - [Open AI Whisper is returning the transcription in English instead of the native language](https://stackoverflow.com/questions/73929778/open-ai-whisper-is-returning-the-transcription-in-english-instead-of-the-native) `stackexchange` — 2022-10-02
  - [How to transcribe Whisper Open AI transcription for large video files larger than 25mb using python?](https://stackoverflow.com/questions/76417335/how-to-transcribe-whisper-open-ai-transcription-for-large-video-files-larger-tha) `stackexchange` — 2023-06-06
  - [Using Whisper AI for transcription from Polish - encoding error?](https://stackoverflow.com/questions/78478908/using-whisper-ai-for-transcription-from-polish-encoding-error) `stackexchange` — 2024-05-14
  - [Enable other languages for meeting AI transcription](https://github.com/calcom/cal.diy/issues/21727) `githubissues` — 2025-06-06
  - [Recurring `TypeError: 'tuple' object has no attribute 'lower'` from `gpt-4o` / `gpt-4o-mini` transcription API (Streaming)](https://community.openai.com/t/recurring-typeerror-tuple-object-has-no-attribute-lower-from-gpt-4o-gpt-4o-mini-transcription-api-streaming/1321082) `discourse` — 2025-07-21
  - [Why does transcription_session.update cause an error? Realtime transcription API bug](https://community.openai.com/t/why-does-transcription-session-update-cause-an-error-realtime-transcription-api-bug/1370560) `discourse` — 2025-12-31
- Decide whether the product is a hosted service, a local app, or a self-hosted API. The signals show all three models already exist, so the wedge has to be use case and reliability, not just deployment style.
  - [Zackriya-Solutions/meetily](https://github.com/Zackriya-Solutions/meetily) `github` — 2024-12-26
  - [pluja/whishper](https://github.com/pluja/whishper) `github` — 2023-08-26
  - [rishikanthc/Scriberr](https://github.com/rishikanthc/Scriberr) `github` — 2024-10-04
  - [Vexa-ai/vexa](https://github.com/Vexa-ai/vexa) `github` — 2025-02-07
  - [TypeWhisper/typewhisper-mac](https://github.com/TypeWhisper/typewhisper-mac) `github` — 2026-02-12
  - [amicalhq/amical](https://github.com/amicalhq/amical) `github` — 2025-05-08

## Competitor strength

| Project | Stars | Downloads/mo | Activity |
|---|---:|---:|---|
| [Zackriya-Solutions/meetily](https://github.com/Zackriya-Solutions/meetily) | 27,920 | — | active |
| [BasedHardware/omi](https://github.com/BasedHardware/omi) | 13,100 | — | active |
| [Anil-matcha/AI-Youtube-Shorts-Generator](https://github.com/Anil-matcha/AI-Youtube-Shorts-Generator) | 4,453 | — | active |
| [tmoroney/auto-subs](https://github.com/tmoroney/auto-subs) | 3,953 | — | active |
| [pluja/whishper](https://github.com/pluja/whishper) | 3,049 | — | active |
| [rishikanthc/Scriberr](https://github.com/rishikanthc/Scriberr) | 2,875 | — | active |
| [Vexa-ai/vexa](https://github.com/Vexa-ai/vexa) | 2,634 | — | active |
| [sindresorhus/awesome-whisper](https://github.com/sindresorhus/awesome-whisper) | 2,364 | — | slow |
| [floneum/kalosm](https://github.com/floneum/kalosm) | 2,215 | — | active |
| [Natively-AI-assistant/natively-cluely-ai-assistant](https://github.com/Natively-AI-assistant/natively-cluely-ai-assistant) | 2,096 | — | active |
| [kaixxx/noScribe](https://github.com/kaixxx/noScribe) | 2,076 | — | active |
| [juanmc2005/diart](https://github.com/juanmc2005/diart) | 2,009 | — | active |

## Domain names

- `transcriptionwhisper.com` — available
- `transcriptionwhisper.io` — available
- `transcription.com` — taken
- `gettranscription.com` — taken

---

Sources searched: hackernews (30), github (20), githubissues (20), discourse (15), intent (2), stackexchange (4), registries (6), tavily (9), devto (10)

Generated by Shows Its Work — https://showsitsworks.com/idea/an-ai-transcription-service
Findings link to their sources. Estimates are marked as estimates.