🎙️
Whisper
by OpenAI
Whisper is OpenAI's open-source automatic speech recognition model, trained on a large enough mix of multilingual audio that it handles accents, background noise, and technical vocabulary noticeably better than the transcription tools that came before it. Because it's free to download and run locally, it's become the engine underneath a huge number of transcription apps and meeting-notes tools rather than something people use directly. Developers who need transcription baked into their own product default to it before looking at paid APIs.
Whisper features and capabilities, including which plan they need and how they compare in the CanIUse matrix.
openaiAdded 2026-07-17Catalog rank 9
Used in combinations
More from OpenAI
ChatGPTAdvanced AI chatbot by OpenAI for conversations and assistanceAdded 2025-09-21chatbottext_generationai_assistantconversationalDALL-E 3OpenAI's text-to-image model inside ChatGPTAdded 2025-01-01image_generationcreativetext_to_imagevisualGPT ImageOpenAI's native image generation model built into ChatGPT and the APIAdded 2026-07-17image_generationcreativetext_to_imagevisualChatGPT CanvasChatGPT's collaborative writing and coding workspaceAdded 2025-09-21writingtext_generationproductivitycontentDaybreakOpenAI's cybersecurity service for defensive security teamsAdded 2026-08-17cloud_basedsecuretrainingmodel_basedOpenAI OperatorOpenAI's web-browsing AI agent that completes tasks autonomouslyAdded 2025-09-21ai_agentautomationautonomousai_assistantsOpenAI PresenceOpenAI's managed platform for enterprise voice and chat AI agentsAdded 2026-07-22ai_agentautomationautonomousai_assistantsOpenAI Realtime APILow-latency speech-to-speech API for voice AI applicationsAdded 2025-09-21developer_toolscode_generationinfrastructuredeployment
Compare Whisper with:
Directory alternatives
Otter.aiAI meeting transcription and note-taking assistantAdded 2025-09-21meetingproductivitytranscriptioncollaborationDescriptAI-powered audio and video editing with transcriptionAdded 2025-09-21audiocreativesoundmusicAssemblyAIEnterprise-grade speech AI API for transcription and audio intelligenceAdded 2025-01-01transcriptionspeech_to_textaudiotextDeepgramReal-time speech recognition API built for productionAdded 2025-01-01transcriptionspeech_to_textaudiotextRevAI transcription and subtitling serviceAdded 2025-09-21video_editingtranscriptionsubtitleaccessibilityFireflies.aiAI notetaker for meetings with transcription and analysisAdded 2025-09-21meetingproductivitytranscriptioncollaborationNottaAI transcription and meeting notes in 58 languagesAdded 2025-09-21transcriptionspeech_to_textaudiotextHappy ScribeAI transcription and subtitle creationAdded 2025-09-21video_editingtranscriptionsubtitleaccessibility
✨ Modalities
💬
Partial. Optional prompt steers spelling/style (224 tokens on hosted Groq/OpenAI-style APIs). The model input is audio.Text Input
🖼️
No. Speech recognition, not vision.Image Upload / Vision
🎨
No. Does not generate images.Image Generation
🎙️
No. Transcribes speech. It does not talk back as a voice assistant.Voice Mode
🎥
No. May transcribe a video’s audio track. It does not watch/understand video.Video Understanding
🎬
No. Not a video generation product.Video Generation
📁
No. Audio files, not PDF/DOCX.File Upload (PDF, DOCX)
💻
No. You run inference. The model does not execute your code.Code Execution
🔊
Yes, requires free plan. Core product: multilingual speech-to-text (and translation to English).Audio Input
🖌️
No. No editor UI.Canvas / Artifact Editor
🖥️
No. No computer-use product.Screen Sharing / Computer Use
🧠 Memory & Context
🧠
No. Each run is a job.Persistent Memory
📏
No. Not an LLM context-window product.Large Context Window
⚙️
Partial. A short prompt can guide spelling. Not a global instruction profile.Custom Instructions
📜
No. Not a chat product.Conversation History
🔌 Integrations
🌐
No. Does not browse the web.Web Search / Browsing
🔌
Partial. Community wraps it in many apps. OpenAI’s repo is the model, not a plugin store.Plugins / Extensions
🔧
Yes, requires plus plan. Hosted by OpenAI, Groq, and others via transcriptions APIs. The GitHub repo is local inference.API Access
⚡
Unknown. No first-party Zapier app was checked.Zapier Integration
🤖
No. Not a bot store.Custom GPTs / Bots
👥
No. No multiplayer product.Real-time Collaboration
🐦
No. No live social firehose.Live Social Data Integration
📚
No. Not a cited-search product.Live Citation Mapping
👨💻 Developer
🛠️
No. Returns a transcript, not tool calls.Function Calling
⚡
No. Typical APIs return the finished transcript (json/text). Not token streaming of a chat.Streaming Responses
🎛️
Partial. Weights are downloadable so the community fine-tunes. OpenAI does not sell a Whisper fine-tune SKU on the GitHub page.Fine-Tuning
📐
No. Not an embeddings API.Embeddings API
📋
Partial. verbose_json with word/segment timestamps. Not arbitrary JSON schema.Structured Output (JSON)
🧩
No. Not an LLM reasoning mode.Reasoning / Thinking Mode
🏠
Yes, requires free plan. Official GitHub repo runs on your hardware.Local Execution
🔓
Yes, requires free plan. Model weights and inference code are public on GitHub (108k+ stars).Open Weights
🤖
No. Not an orchestrator.Multi-Agent Orchestration
📱 Platforms
🌐
No. No official Whisper web app. Hosted playgrounds are third-party (e.g. Groq).Web App
📱
No. No official iOS app.iOS App
🤖
No. No official Android app.Android App
🖥️
No. CLI/Python package, not a desktop app.Desktop App
🔵
No. No official Chrome extension.Chrome Extension
🔒 Privacy
🔒
Unknown. Local runs stay on your machine. Hosted APIs need that vendor’s DPA.GDPR Compliant
🚫
Unknown. Local inference sends nowhere. Hosted API training policies vary by vendor.Opt Out of Data Training
🏢
Partial. Self-host for isolation. No Whisper Enterprise SKU on GitHub.Enterprise Privacy Mode
💼 Workspaces
📂
No. No project product.Projects / Folders