One-tap send for iPhone dictation

[shivamhwp]

When I dictate on the iPhone app, sending takes two taps. I tap the checkmark, wait for the transcript, then tap send. Most of the time I don't touch the text, so that second tap is just me waiting.

I know the review step is on purpose. The voice input notes say transcription only edits the draft and never sends. I'm not asking to change the default. I'm asking if there's room for an opt-in way to skip it.

Two ways it could work:

1. A send button next to the checkmark while recording. The checkmark inserts the text like it does now. Send transcribes and sends in one tap. X still cancels.
2. A "Send dictation immediately" setting, off by default. With it on, the checkmark sends instead of inserting.

I'd go with the button. You choose per message, and nothing changes for anyone who doesn't tap it. A setting works too if you'd rather keep the toolbar as it is.

Either way, it should never send when nothing usable came back: no speech, a draft that changed during transcription, a cancel, or an error. If the send is blocked, say uploads are still running, the text just stays in the box.

Earlier attempts:

• feat: add web voice input and optional auto-send #12893 was the setting version, for iPhone and web. It was closed because it also added web dictation without approval, not because of the auto-send part.
• feat(mobile): send a dictated message in one tap #14712 is the button version, iPhone only. It was closed under the prior-approval rule. The code is written and CI passed, so if you approve either option I can reopen it or switch it to the setting.

Would you take either of these? If so, which one?

[Dwite]

I opened #14807 for this. It is your option 2: a Send immediately setting for iPhone voice input, off by default.

With the setting on, the checkmark sends the whole composer message through the normal send path. If the send is blocked, the text stays in the composer as a draft.

It is waiting for review and is not approved yet, so it may still change.

[gitkamaal]

+1 for option 2 (an opt-in "Send immediately" setting). I'd also like the same setting on desktop once composer voice input lands there (#14814), so dictation behaves the same on every client. Most of my dictated prompts go out unedited, so the extra Enter or tap is just friction.

[goleary]

+1, and I'd vote for option 1 (a send button next to the checkmark while recording). Other voice apps I use, such as Muse, do this: tap the mic, talk, then tap an up-arrow that transcribes and sends in one step. Choosing per message is better than a global setting. When I want to check the transcript I still tap the checkmark, and when I don't, it's one tap. On iPhone, most of my dictated prompts go out unedited.