Review and insert dictated text
The transcript goes into the prompt field, not into your content. This page covers exactly where it lands and what to do with it.
Where the text goes
When transcription succeeds, the text is appended to the prompt field the control belongs to. Three details follow from that:
Nothing is overwritten. Text you already typed stays. The transcript is added after it.
A line break is inserted when needed. If the field already ends with whitespace the transcript is joined directly; otherwise a newline is added first, so a dictated sentence never runs into the end of a typed one.
The cursor is placed at the end. You can carry on typing immediately without clicking back into the field.
The status line confirms it:
Transcription added to the prompt.
Dictating in several passes
Because the transcript appends, you can build a prompt in pieces: dictate a paragraph, type a correction, dictate another paragraph.
Each recording is charged separately, at six credits per started minute, so this is convenient but not free. Two thirty-second recordings cost twelve credits; the same content in one recording costs six. Dictate in longer passes when you can.
Always read it before generating
The transcript is inserted without review. Read it before you press a generate button, because a transcription error becomes an instruction.
Look for:
- Names and terms. Product names, brand names and industry vocabulary are the most likely errors. See Dictation language and punctuation.
- Numbers. Prices, quantities, dates and percentages are frequently wrong, and they are exactly the values a generator will repeat confidently.
- Sentence boundaries. Missing or misplaced punctuation can change what an instruction means.
- Your own thinking-out-loud. Phrases like "no, actually" and "let me start again" are transcribed faithfully. Delete them, because the generator will treat them as part of the brief.
That last one is the most common problem with dictated prompts, and the easiest to fix.
Edit before you generate, not after
The prompt field is an ordinary textarea. Editing it costs nothing; regenerating because the prompt was wrong costs credits.
A useful sequence:
- Dictate the rough version.
- Delete the false starts and repairs.
- Tighten it into instructions rather than narration.
- Check the character counter below the field.
- Then estimate and generate.
Dictation produces spoken-shaped text — longer and looser than written instruction. Trimming it usually improves the result and lowers the cost, since the prompt is part of what an operation is priced on.
Character limits still apply
Every prompt field has a limit — 100,000 characters in the long-form post builder, 3,100 in the image tools — with a live counter below the field.
A three-minute dictation can approach those limits. If the counter shows you are over, the tool will not submit until you cut it back. Trim the narration first; it is usually where the excess is.
Combining dictation with templates
Prompt templates also append to the field. You can apply a template and then dictate the specifics, or dictate first and then apply a template that adds your standard exclusions.
Order matters only for readability, since both append. Applying the template first and dictating into the gaps tends to read better. See Use saved prompt templates.
If the transcript is empty or wrong
Nothing is inserted when the transcript comes back empty, and the status reports it. The recording was still charged if the service processed it — see Troubleshoot dictation.
If the transcript is present but badly wrong, the usual causes are microphone distance, background noise or a mismatch between the language you spoke and the site language. Fixing the input beats re-recording repeatedly.
What it never does
The transcript is never written into a post, a page, an image caption or any other content. It only ever fills the prompt field you were working in.
Dictation is an input method. Everything that happens afterwards is the ordinary tool workflow, with its own estimate and confirmation.