تبدیل گفتار به متن
Superscribe
Turns calls on your existing number into CRM-ready notes, with desktop dictation and auto-matched billable time
نتیجه ارزیابی
بخش تایپ صوتی روی دسکتاپ در یک نشست قابل پیادهسازی است، اما محصول اصلی فراتر از این است: رونویسی زنده تماسها روی خط اصلی آیفون نیاز به دایورت، شمارههای Twilio، نرمافزار CallKit/PushKit و خطوط صوتی دارد. علاوه بر این، حدود ۴۵ هزار خط کد API و وباپ برای انتساب زمان تماسها به پروژههای مختلف (همگام با گیتهاب)، گزارشها، سرچ معنایی و سرور MCP در کار است که پیادهسازی آن چند هفته زمان میبرد.
- زمان تقریبی ساخت
- یک نشست (فقط تایپ صوتی)، چند هفته در صورت ضبط تماس
- هزینه تقریبی اجرا
- نامشخص
- اشتراک محصول اصلی
- ۳۸ دلار در ماه
چه چیزی را از دست میدهی؟
- ضبط مکالمات روی شماره اصلی گوشی (دایورت خط به سمت یک سافتفون با CallKit)
- تایپ زنده و روان کلمه به کلمه هماهنگ با اپهای مختلف: ترمینال، ویرایشگرهای الکترون و ادیتورهای وب
- دستهبندی خودکار: تطبیق تماسها و دیکتهها با پروژهها از طریق Embeddingها و کامیتهای زنده GitHub جهت ثبت دقیق ساعات کارفرما بدون نیاز به تگ زدن دستی
- محیط کاری جامع: جستجوی مفهومی، گزارشهای PDF فاکتور، پیشنویس یادداشتهای CRM، یک API عمومی و سرور MCP
- سازگاری با ویندوز و فایلهای نصب امضاشده و رسمی با قابلیت آپدیت خودکار
پرامپت شروع
برای شروع ساخت، مستقیم کپیاش کن.
Build me a push-to-talk dictation tool for macOS to replace Superscribe's desktop app. Requirements: - Swift menu bar app, SPM only, no Xcode project. A global hotkey (default: hold Option+Space) records while held, stops on release. - Capture the mic with AVAudioEngine, downsample to 16kHz mono PCM, and stream it over WebSocket to ElevenLabs Scribe realtime (key in .env). Show partial transcripts in a small floating panel while I speak. - On release, run one LLM cleanup pass over the final transcript (fix punctuation, drop filler words; key in the same .env), then paste it into the focused app via NSPasteboard + CGEvent Cmd-V and restore my previous clipboard afterwards. - Fallback: with no ElevenLabs key, record to a temp wav and transcribe locally with whisper.cpp instead. Slower is fine. - Menu bar icon shows idle/recording/transcribing states; a history window lists the last 20 transcripts with copy buttons, persisted to ~/Dictation/history.jsonl. - Ad-hoc codesign for my own machine only. README documents the Microphone and Accessibility permission prompts and the whisper.cpp model download. - Out of scope: live word-by-word typing into the field (paste on release only), Windows support, accounts and billing, and the phone-call capture product. If I want call notes I will upload recordings by hand.
برای ساختش کمک میخوای؟
ابزارهای مشابه
اگر این ابزار را بررسی میکنی، این گزینهها هم احتمالاً به کارت میآیند.
به درد یکی دیگه هم میخوره؟




