Who whishper is for#
Journalists and researchers handling sensitive interviews
Whishper runs entirely on your hardware, so interview audio never leaves your infrastructure. Combine URL-based or file-based transcription with the built-in subtitle editor to produce timestamped text for fact-checking or archival.
Skip if:
The audio sources are not sensitive and you want the convenience of a managed API with no server overhead. In that case, Otter.ai's managed tier is simpler to start with.
Content creators building subtitled video
Upload a video file or paste a YouTube URL, let Whishper transcribe it with FasterWhisper, and export an SRT or VTT subtitle file ready for your video editor. The CPS warnings in the subtitle editor flag over-paced captions before export.
Skip if:
You need automated subtitle burning (rendering captions into the video file). Whishper exports subtitle files for external editors; it does not render video with burned-in subtitles.
Developers integrating a self-hosted transcription backend
Whishper exposes a Transcription API layer that wraps FasterWhisper. Teams building internal tools can self-host Whishper as the transcription backend and call it from their own applications.
Skip if:
You need a fully documented REST API with official client libraries. Whishper's API is for internal use and its documentation is limited; calling faster-whisper directly may serve production needs better.
Privacy-conscious teams processing meeting recordings
Teams that record internal meetings but cannot use cloud services due to compliance or policy constraints can use Whishper to generate transcripts and translate them for team members, all on internal infrastructure.
Skip if:
Your team needs speaker diarization (identifying who said what). Whishper does not currently support speaker identification.
The problem it solves#
Cloud transcription services charge per minute, per hour, or per seat, and they process your audio on someone else's servers. For researchers handling confidential interviews, journalists protecting sources, medical professionals dealing with patient data, or developers building transcription-dependent tools, that trade-off is unacceptable. Even for lower-stakes use cases, the costs add up fast: a team transcribing hours of meeting recordings per week quickly exceeds what flat-rate subscriptions cover.
The other problem is output control. Most commercial transcription tools lock subtitles and transcripts into their own editors and export formats. When you need SRT files for a video editor, VTT for a web player, or plain text for a search index, you are dependent on the vendor's export pipeline. A self-hosted transcription suite eliminates the cost model and the data dependency together.
How it solves it#
100% local processing, zero cloud dependency
Every transcription, translation, and subtitle edit runs on your machine. No audio is sent to external servers, and the tool can work fully offline after initial setup. Your files stay yours.
FasterWhisper transcription engine
Uses FasterWhisper as the Whisper backend, which delivers significantly faster transcription on CPU compared to the original OpenAI Whisper model. NVIDIA GPU acceleration is also supported for even faster results on compatible hardware.
Built-in subtitle editor
Edit subtitles directly in the web UI without exporting first. The editor supports CPS (characters per second) warnings to flag over-paced captions, segment splitting, segment insertion, and real-time highlighting based on media position.
60+ language translation
Translate transcriptions to and from over 60 languages using Argos Translate and LibreTranslate. Translation runs locally in a separate container, so no external API is needed for multilingual output.
Multi-format subtitle export
Download transcriptions in TXT, JSON, VTT, or SRT formats, or copy raw text to the clipboard. VTT and SRT exports are ready for direct use in most video editors and web video players.
URL transcription via yt-dlp
Transcribe media directly from a URL without downloading the file first. Any source supported by yt-dlp (YouTube, Vimeo, and hundreds of other platforms) can be queued by URL in the web UI.
Strengths and trade-offs#
Strengths
- Complete data privacy by designAll processing happens locally. Whisper model inference, LibreTranslate translation, and subtitle editing never call external APIs. For teams under data protection requirements (legal, medical, journalism), this is the key differentiator from cloud services like Otter.ai or Rev.com.
- No per-minute or per-seat pricingCloud transcription services charge by usage. Rev.com charges per minute for human transcription; Otter.ai charges per seat monthly. Whishper has no usage-based pricing: run as many transcriptions as your hardware allows.
- Works on CPU, no GPU requiredFasterWhisper's efficiency means Whishper runs acceptably fast even on CPU-only servers. GPU is optional and provides additional speed if NVIDIA hardware is available, but it is not a requirement for getting started.
- All-in-one transcription and subtitling workflowMost open source transcription tools are CLI-only or output raw text. Whishper adds a full subtitle editor in the web UI, covering the complete workflow from upload or URL to export without switching tools.
Trade-offs
- -Main branch frozen: v4 rewrite in progressThe README notes that the current main branch will not receive new releases or updates while a full v4 rewrite is underway on a separate branch. This means bug fixes and new features will not land in the current version until v4 ships. For teams evaluating stability, this is an important consideration.
- -No user authenticationUser authentication is listed on the project roadmap but not yet implemented. Anyone with network access to the deployed instance can use it. Running Whishper on a public-facing server without additional access controls (reverse proxy auth, VPN, firewall rules) is not advisable.
- -Multi-container self-hosting setupWhishper runs as six services: a transcription API, a Go backend, a SvelteKit frontend, LibreTranslate, MongoDB, and Nginx. This is not a single-container install. Operators need Docker Compose experience and enough server resources to run all six containers concurrently.
- -NVIDIA GPU only for hardware accelerationGPU acceleration is limited to NVIDIA hardware. AMD GPU support is listed as a roadmap question without a clear timeline. Apple Silicon (M-series) GPU acceleration is not mentioned in the docs.
whishper vs alternatives#
Whishper vs Otter.ai
Otter.ai and Whishper both convert spoken audio to text, but they serve opposite deployment models. Otter.ai is a managed cloud service; Whishper is a self-hosted suite that runs entirely on your own infrastructure.
| Feature | Whishper | Otter.ai |
|---|---|---|
| License | AGPL-3.0 | Proprietary |
| Deployment | Self-hosted (Docker) | Cloud only |
| Data privacy | Local processing | Processed on Otter.ai servers |
| Pricing | Free (self-hosted) | Per-seat monthly subscription |
| Translation | 60+ languages (local) | Limited |
| Subtitle editor | Yes, built-in | No |
Whishper is the better choice when audio privacy is a hard requirement, when you process high volumes that would exceed managed plan limits, or when you need subtitle editing in the same workflow. Otter.ai is the better choice for teams that want a managed service, need calendar integration and meeting bot features, or cannot maintain a self-hosted server.
Whishper vs Descript
Descript is a video and podcast editing tool with AI transcription built in. Whishper is a transcription-first tool with subtitle editing; it does not edit audio or video files.
| Feature | Whishper | Descript |
|---|---|---|
| License | AGPL-3.0 | Proprietary |
| Deployment | Self-hosted (Docker) | Cloud only |
| Data privacy | Local processing | Processed on Descript servers |
| Subtitle editor | Yes | Yes |
| Video editing | No | Yes |
| Pricing | Free (self-hosted) | Per-seat monthly subscription |
Whishper is the better choice when you only need transcription and subtitle export, when audio privacy matters, or when cost control is the priority. Descript is the better choice when you need a complete audio and video editing environment with transcription built in.
Whishper vs Rev.com
Rev.com offers both AI and human transcription as a managed service, billed per minute. Whishper provides AI transcription locally at no per-minute cost.
| Feature | Whishper | Rev.com |
|---|---|---|
| License | AGPL-3.0 | Proprietary |
| Deployment | Self-hosted (Docker) | Cloud only |
| Data privacy | Local processing | Processed on Rev.com servers |
| Pricing | Free (self-hosted) | Per-minute billing |
| Human transcription | No | Yes |
| Translation | 60+ languages (local) | Limited |
Whishper is the better choice for teams with predictable, high-volume transcription needs where per-minute billing adds up. Rev.com is the better choice when you need human-reviewed transcripts for legal or compliance purposes, or when one-off transcription is more practical than running a self-hosted stack.
What it's built on#
- Languages
- GoJavaScriptPython
- Frameworks
- FastAPISvelte
FAQ#
Does whishper send my audio to external servers?
No. Every component runs locally: FasterWhisper handles transcription, LibreTranslate handles translation, and MongoDB stores your transcription history. None of your audio or text leaves your machine unless you configure it to.
Can whishper run without a GPU?
Yes. FasterWhisper is efficient enough to transcribe on CPU-only hardware. GPU acceleration (NVIDIA only) is optional and speeds up processing but is not required to run the tool.
What file formats does whishper support for export?
Whishper exports transcriptions in TXT, JSON, VTT, and SRT formats. You can also copy raw text directly to your clipboard from the web UI.
Is whishper still actively maintained?
The main branch is currently frozen while a full v4 rewrite is underway on a separate branch. The project received its last push in July 2026. Check the v4 branch on GitHub for active development progress.
How many languages can whishper translate transcriptions into?
Whishper supports translation to and from over 60 languages using Argos Translate and LibreTranslate. Both translation libraries run locally in a Docker container, so no external translation API is needed.
Similar open-source tools#
auto-subs
Local-first AI subtitles for DaVinci Resolve and Premiere Pro.
Handy
Free offline speech-to-text: runs on Windows, macOS, and Linux
FluidVoice
Voice-to-text for macOS with on-device AI and zero cloud dependency
omi
Open AI memory assistant for meetings, screens, and wearables
Ghost Pepper
Local voice dictation and meeting transcription for macOS
CLI-Anything
Empower AI agents with agent-native CLIs