The ready-to-use OpenAI Whisper alternative
Get timestamped transcripts with summaries, exports and speaker labels – no Python, PyTorch or GPU to manage. Upload a file or paste a link, and buy credits that never expire.
60 min free · no card required · we do not train our own models on your content
Free without chopping up the recording
Files up to 25 GB on a free account
Keep a full interview, lecture, podcast, or video in one piece. Every account, free or paid, takes files up to 25 GB, and recordings can be any length. Your first 60 minutes are a one-time allowance, and no card is needed.
Is Pepys a good OpenAI Whisper alternative?
Pepys is a ready-to-use alternative to OpenAI Whisper for anyone who'd rather not self-host. Whisper is an open-source model you run yourself with Python, PyTorch and ideally a GPU; Pepys is a web app with summaries and exports built in, and speaker labels (diarization) at 1.5 credits per minute. Buy credits once and they never expire. 60 minutes free, no card.

OpenAI Whisper is one of the most accurate open speech models you can self-host. The large-v3 checkpoint and the faster large-v3-turbo variant are both MIT-licensed for code and weights, so you can run them almost anywhere. The catch is that Whisper is a model, not an app: you install Python, PyTorch and ffmpeg, drive it from the CLI, and ideally point it at a GPU. Diarization isn't built in, so speaker labels mean bolting on a separate tool like pyannote. And the hosted API caps uploads at 25 MB per request.
Pepys is what you reach for when you just want the transcript. You get the same kind of accurate, timestamped output across ~99 languages, but as a web app with AI summaries, chapters, chat and one-click exports already built in, and speaker labels at 1.5 credits per minute. Upload a file or paste a YouTube or podcast link – long files are chunked and stitched automatically, so there is no 25 MB request limit: every account, free or paid, takes files up to 25 GB, and recordings can be any length. You buy credits as you need them, they never expire, and we do not train our own models on your content. Start with 60 minutes free, no card.
Pepys vs OpenAI Whisper, side by side
OpenAI Whisper vs Pepys, head to head
Public, verifiable facts only. Where something isn't publicly stated, we leave it blank rather than guess.
| Feature | Pepys | OpenAI Whisper |
|---|---|---|
| What it is | Ready-to-use web app – nothing to install | Open-source speech-to-text model, plus OpenAI's hosted API |
| Setup required | None – sign up and transcribe in your browser | Self-host: Python 3.8–3.11, PyTorch and ffmpeg (rust may also be needed) |
| Consumer app / interface | Full web app, works on any device (no install) | No official GUI – CLI and Python only (open-source) |
| Hardware / GPU | None – runs in the cloud on any device | Runs on CPU; a GPU is recommended for speed |
| Pricing model | Pay-as-you-go credits, bought once – they never expire | Open-source model free to self-host (MIT); hosted API billed per minute/token |
| Cost per minute | About $1/hour (from $0.85/hr at volume), everything included | Self-host: free. API: whisper-1 $0.006/min; gpt-4o-mini-transcribe ~$0.003/min (est., token-billed) |
| Speaker labels / diarization | Yes – speaker labels at 1.5× credits | Not built into the open-source model (add pyannote); OpenAI's hosted API now offers built-in diarization |
| Languages | 99, auto-detected | ~99 (English plus 98 other languages) |
| Paste a link to transcribe | Yes – YouTube, podcasts, social, and more | No – local audio files only, no built-in URL / YouTube ingestion |
| Max file size per upload | Files up to 25 GB on any account; recordings of any length, long files chunked automatically | Hosted API: 25 MB per request; self-host limited by your own hardware |
| Built-in AI summaries & chat | Yes – summaries, chapters, and chat with the transcript | Transcription only |
| Trains AI on your data? | No – we do not train our own models on your content | Self-host: stays on your machine. API: not used for training by default (opt-in only) |
Why people switch to Pepys
No setup – there's nothing to install. No Python, PyTorch, ffmpeg or GPU; you sign up and transcribe in your browser on any device.
Summaries, chapters and chat come built in, and diarization costs 1.5 credits per minute. With open-source Whisper you'd assemble those from separate tools like pyannote.
Paste a YouTube, podcast or social link, or upload files of any duration – long files are chunked and stitched automatically, with no 25 MB per-request limit.
Buy credits that never expire, and we do not train our own models on your content.
Any language – 99 detected automatically
- English
- 中文
- Español
- العربية
- हिन्दी
- Français
- 日本語
- Português
- Русский
- Deutsch
- 한국어
- Italiano
- বাংলা
- Türkçe
- فارسی
- Tiếng Việt
- தமிழ்
- Polski
- ไทย
- Українська
- Nederlands
- עברית
- Ελληνικά
- తెలుగు
- Bahasa Indonesia
- اردو
- Svenska
- मराठी
- Română
- Magyar
- Čeština
- ગુજરાતી
- Kiswahili
- ქართული
- Tagalog
- አማርኛ
Common questions about OpenAI Whisper and Pepys
Is Pepys better than OpenAI Whisper?
For most people who'd rather not run Python, PyTorch and a GPU, yes – Pepys gives you AI summaries, one-click exports and speaker labels in the browser, with pay-as-you-go credits that never expire. But open-source Whisper is genuinely the better choice when you want a free, MIT-licensed model to fine-tune, or to run fully offline on your own hardware.
Do I need a GPU or any coding to use Pepys?
No. Whisper's open-source models run best with Python, PyTorch, ffmpeg and ideally a GPU. Pepys is a web app – you sign up, upload a file or paste a link, and transcription runs in the cloud. It works on any device with no install and nothing to configure.
Can Pepys label speakers like Whisper plus pyannote?
Yes, on any signed-in account, free minutes included. Open-source Whisper doesn't diarize on its own, so speaker labels usually mean adding a separate tool like pyannote. Pepys separates the voices automatically as Speaker 1, Speaker 2 and so on (1.5 credits per minute), which you can rename, alongside timestamps, summaries and exports, with no extra pipeline to assemble.
Is the OpenAI Whisper API cheaper than Pepys?
Per minute, often yes – whisper-1 is about $0.006 a minute, and gpt-4o-mini-transcribe works out to roughly $0.003 a minute (an estimate, since it's billed per token). But that's the raw model only, with a 25 MB per-request limit and no app around it. Pepys bundles summaries, link ingestion and exports, with speaker labels at 1.5 credits per minute, from about $1 an hour, and your credits never expire.
Can I export transcripts in the same formats as Whisper?
Yes, and more. Whisper's CLI writes TXT, VTT, SRT, TSV and JSON. Pepys exports TXT, Markdown, DOCX, PDF, SRT, VTT and open JSON, so subtitles, documents and developer-friendly output are all one click away.
Is it free to try?
Yes. Your first 60 minutes are free with no card required, so you can compare Pepys' transcripts against your own Whisper output before you pay. After that you buy credits once – 1 credit is 1 minute, and they never expire.
Compare more alternatives
Popular tools
Don't just take our word for it.
Ask ChatGPT, Claude, or Perplexity how Pepys stacks up against OpenAI Whisper.
Try Pepys free – no card
Pay as you go – credits never expire, nothing to cancel. Compare the transcript against OpenAI Whisper with 60 free minutes.