Spokenly

@Cloud_san99 原因はAIプロンプト機能です。現在Spokenly側のAIサーバーを使っているため、PROが必要になっています。自分のAIキー（OpenAI等）を設定するか、プロンプトを削除すれば解決します。メールでも詳細お送りしました！

Spokenly ✕ Soniox APIでなぜか音声テキスト変換後にアプリが立ち上がって画像のようなポップアップが出るようになった。 アップデートしても一緒だし・・・。 とりあえず原因問い合わせてます📨

Spokenly for Mac is now available outside the App Store. spokenly.app/download or brew install spokenly

Setup guide: spokenly.app/docs/ios/clipb… Works with Action Button and Back Tap.

🎉 Spokenly iOS v1.6.4 is available! New: Clipboard Dictation Dictate to your clipboard without opening the app. Press the Action Button, speak, paste. Done.

@kinsonwu08 AI Prompts for iOS are nearly ready and should launch next week. They’ll include all BYOK options, just like the Mac version.

@SpokenlyApp can the iOS version customize the AI prompt? It is a useful function in MacOs.

🎉 iOS version is here! Spokenly for iOS brings fast audio transcription to your iPhone. All models from the macOS version are supported, including BYOK and local models. Download: apps.apple.com/app/id67403155…

🎉 Spokenly v1.1.0 is available on iOS! Now you can use voice input across all iOS apps. apps.apple.com/app/id67403155…

@repostingg For now, the cancel shortcut can't be changed or disabled, but this is a popular request and will be added in upcoming updates!

@SpokenlyApp Hello! I have a question about the app. I've tested it a few times to record and transcribe my meetings, but every time I pressed the ESC key, the recording was canceled. Is it possible to remove or change the cancellation shortcut?

🚀 Spokenly is live on Product Hunt today. Your upvote would mean a lot: producthunt.com/products/spoke… ✅ Type 4x faster with your voice. ✅ Local voice-to-text models keep your data private and free. ✅ Works seamlessly with Cursor, ChatGPT, or any app where you type.

🚀 Spokenly 2.12.0 is out! Turn spoken words into Mac actions. Search Google, ask ChatGPT, launch apps, run Apple Shortcuts, and more, all through voice commands. spokenly.app/help/agent_mode apps.apple.com/app/id67403155…

@ch1y1z1 @arkuy99 感谢反馈！我会在即将发布的更新中加入对兼容 OpenAI 服务的实时连接支持。

@SpokenlyApp @arkuy99 可以参考macwhisper的兼容的openai接口，会优先使用实时的websocket的接口，体感很快

谁写一个 mac 上语音输入 vibecoding 的 ， 俺原意付费20刀购买

@ch1y1z1 @arkuy99 目前应用还不支持实时转录：它会先完整录音，然后将文件发送到 Speaches 进行文字转换。如果 Speaches 支持实时流式转录，我可以把这个功能加入进来。请告诉我，这对你是否有帮助？

@SpokenlyApp @arkuy99 顺便问一下这个是实时转录的吗

@arkuy99 @ch1y1z1 感谢反馈！您使用的是哪种模型？识别质量取决于模型，建议尝试 GPT-4o。

@SpokenlyApp @ch1y1z1 交互还可以 中英混合识别不准

@ch1y1z1 @arkuy99 你好！Spokenly 允许直接与本地部署、兼容 OpenAI 的端点协同工作，包括 Speaches。查看语音模型的 API 部分。如有疑问，欢迎联系！

@arkuy99 因为自部署的这个实时性相当好，使用 realtime 模型，支持边说边转录，并且是在我自己服务器上部署的，模型不卸载，可以做到实时响应，语音输入任务上基本 说完的一瞬间 字就可以全部蹦出来了。我用其他产品基本都没有这么好的实时性，我用过 groq 、elevenlab 的 api ，spokenly ，都不如自部署的快。

@clyde_yi @AI_vs_ERICK Thank you for reporting this! I’ll work to improve rendering in the upcoming update. The issue likely occurs because the blur and glow effects are not yet optimized

@SpokenlyApp @AI_vs_ERICK This is a great app, but I've noticed that the GPU usage is particularly high during recording, leading to some slight heating, even when using online models and APIs.

@SpokenlyApp Hey, can I ask you what's the best model that gives the best performance that you've tried? I only have a MacBook M2 Air, so I've been running the online large B3 Turbo. Is that the best one? Recently, I noticed there is some latency, especially compared to the Wispr Flow that other proprietary app model that costs $15/month. I even bought a new headset with a better mic just to control for that.

@AI_vs_ERICK By the way, I just set up Discord for the app, feel free to send your feedback there if you prefer: discord.gg/DgUeuway

@AI_vs_ERICK Just completed the first step of migrating to the new hosting provider. Performance seems significantly faster now, even with GPT-4o models, although it may vary depending on your location. Let me know if you notice any changes on your end. Thanks!

@AI_vs_ERICK This is now live in the latest version of the app, thanks!

@SpokenlyApp maybe drag and drop reordering would be good? For using them with the command number keys to select them.

@AI_vs_ERICK I usually go with the GPT-4o model because it’s more accurate and handles punctuation better. It’s slower, but I end up saving time since I don’t have to fix mistakes afterward

@AI_vs_ERICK Hey! Thanks for reporting this. Other users have mentioned it too. Right now I’m migrating to a new hosting provider, and the issue should be resolved in a couple of days. For faster but slightly less accurate results, you can try the Real-time Nova 3 model