Useful words. Thoughtfully arranged.

Your first Mac recording shortcut

Control means the key labelled control or ctrl. It is not Command or Option.

Choose the correct Mac edition

Look for Desktop · Full Mac when you need the complete Mac controls.

Set up microphone and Accessibility

Only you can approve macOS permission prompts. Sayloom cannot silently grant access.

Choose a destination before recording

Changing the field while recording can deliberately stop insertion to protect existing text.

Begin with English dictation

English Small is used for the English automatic path. A text model is not required.

Make your first mobile note

Mobile uses its platform speech service. Availability differs by language and device.

Set up a writing workflow on iPad

Do not assume the Mac global shortcut or Mac Ollama runtime exists on iPad.

Understand the ten-use trial

Ten is the total allowance. Earlier successful uses remain counted after upgrading.

Review a transcript before sharing

Automatic insertion is not automatic approval to send a message.

Back up the workspace before changes

Workspace exports exclude audio, downloaded models and stored API keys.

Why a full sentence helps a first test

A short hello is a weaker test than a complete sentence with a known expected result.

Speak a new paragraph

English voice line commands must be enabled. Review the resulting paragraph break.

Speak a new line

New line is a formatting command; do not assume every editing command is implemented.

Handle names and project vocabulary

Vocabulary hints can help but do not guarantee correct spelling.

Check numbers after dictation

A model can render spoken numbers as digits or words. Confirm meaning rather than style alone.

Pause naturally without losing your thought

Stop is an explicit action. Do not depend on silence to finish the recording.

Switch from English to multilingual speech

English-only model files cannot become multilingual merely by changing a label.

Dictate into a Sayloom text field

Focus an editable native field before using the recording shortcut.

Replace a selected passage safely

Check the selection before starting. Do not edit the destination while recognition is running.

Dictation does not require a writing model

Ollama 1B and 3B are optional text models, not microphone or speech models.

Give an email a readable greeting

The name and signature came from the speaker; the app should not invent them.

Confirm a meeting without changing the time

Check that formatting did not turn a request into a confirmation.

Write a concise follow-up

A follow-up question must remain a question in the final draft.

Decline a time while preserving your intent

Negation is important. Read does not carefully before sending.

Dictate a project status update

Do not let a polished style make unfinished work sound complete.

Keep an email question intact

Small writing models can answer a question instead of formatting it.

Supply your own sign-off

The app cannot know which name should sign an email unless you provide it.

Use local Email layout when AI is unavailable

A provider error should not erase the source transcript.

Keep a longer email readable

Check every action item against the original; a shorter draft can omit a detail.

Compare an email before inserting it

Review and approve are different actions, even if a model treats them as similar.

Keep project names consistent

This example assumes a replacement from say loom to Sayloom exists.

Expand a spoken booking link

Replace the sample URL with your own before saving the rule.

Preserve the acronym your team uses

Only abbreviate when that exact replacement is appropriate in your work.

Understand longest matching phrases

Overlapping rules should not cascade through newly inserted text.

Avoid changing words inside other words

A whole-word rule for ink should leave thinking alone.

Keep Unicode text in a replacement

Check combining marks, non-Latin scripts and punctuation in the actual destination.

Use vocabulary hints for unusual words

Hints are not guaranteed spellings and should not replace output review.

Merge a dictionary carefully

Review imported rules. Matching spoken phrases can update existing entries.

Export a portable vocabulary backup

A dictionary export is not a backup of API keys, model files or recordings.

Review correction suggestions before adding them

Accepting a correction is a deliberate choice; unrelated rewrites are not useful dictionary entries.

Choose between Ollama 1B and 3B

Existing Sayloom sample checks found instruction-following problems in 1B. That is not a universal benchmark.

Separate speech models from writing models

Downloading an Ollama text model does not enable microphone capture.

Use the included English speech model

English-only recognition is a different model choice from multilingual detection.

Use the multilingual speech path

Do not promise every language on every device; model and platform support vary.

Understand the local Ollama setup

Install Ollama from its official source; model downloads need local storage and memory.

Think about model memory and latency

Download size, runtime memory and time to respond are different measurements.

Configure an optional text-provider key

Keys belong in the app’s configured storage, never in a public website or screenshot.

Why Mac model files do not run on phones

The mobile implementation uses native speech and an optional HTTPS text provider.

Understand optional local text prewarming

Prewarming can trade memory or background work for latency. Measure the actual effect.

Switch writing models without deleting them

Choosing a model and deleting its files are separate actions. Keep both models if you still want to compare.

Choose the lightest writing mode

A simple dictation does not need an assistant answer or an AI rewrite.

Add paragraph breaks deliberately

English voice commands must be enabled before spoken breaks are interpreted.

Remove standalone vocal fillers carefully

The optional cleanup targets standalone English fillers, not every hesitation in every language.

Keep the original while improving a draft

An original transcript is the reference when a model makes an unexpected change.

Keep a dictated question as a question

A model can respond to a question instead of rewriting it. Compare intent before sending.

Protect names and dates during formatting

Changing a person or date is not a harmless stylistic edit.

Preview an optional AI transformation

Provider output is not proof that the microphone recognized the source correctly.

Use selected text as explicit context

Context capture depends on permissions and the writing app’s exposed text.

Keep a timestamped Markdown journal

Confirm the journal destination before appending a note.

Choose paste, copy or response output

A command output can execute a configured workflow. Enable commands intentionally.

Follow the local speech path

Local speech is a property of the selected path, not every optional provider.

Understand optional cloud speech consent

Cloud speech transmits audio to your configured HTTPS provider.

Choose only the context a mode needs

A clipboard or window can contain information beyond the sentence you just spoke.

Decide whether to retain recording audio

Completed audio retention is optional; imported files have their own original storage location.

Keep provider keys out of screenshots

Redact credentials before sharing logs, settings images or workspace files.

Know what a workspace backup includes

Workspace exports include transcript text but exclude audio, downloaded models and stored secrets.

Why password fields stay protected

The app deliberately rejects secure inputs. Do not disable that protection to force a test.

Understand the limits of an offline trial

Uninstalling or clearing device data can affect offline trial protection.

Share useful diagnostics without private text

Use permission and runtime metadata first. Personal recordings are not required for every support case.

Separate store access from provider charges

Lifetime access does not include hosted AI usage, and Apple and Google purchases are separate.

Capture a fleeting idea in Notes

Check the field is editable before starting the Mac shortcut.

Dictate into a browser form

Browser custom editors expose different text APIs. Clipboard fallback may work when direct insertion does not.

Draft a chat message without sending it

Normal recording stop should not automatically send your chat message.

Speak a prompt for another AI app

Dictation captures the prompt; the destination AI app is a separate service.

Keep a simple daily project journal

Check the configured Markdown file path before using Journal output.

Turn an audio file into a draft

Import only recordings you are entitled to process and review long files in manageable sections.

Retry retained audio with another setup

Retry requires retained audio. Turning retention on later cannot recover a discarded recording.

Use the Android voice keyboard

Android app and keyboard share trial usage. Device speech support still matters.

Use the iOS opening shortcut

The implemented iOS shortcut opens Sayloom; it is not a Mac-style background global recorder.

Search inside Sayloom with your voice

Focus the History search field before recording. A search with no matches can still be inserted correctly.

Check recognition quality with a known sentence

One successful sample is useful evidence, not a guarantee of accuracy for every speaker or environment.

When Dictation says no clear speech

Check input selection and timing before changing an unrelated text model.

Check a quiet microphone signal

Speak after Listening. Check macOS input level and the actual microphone in use.

When English produces another script

Choose English and its included English model before comparing multilingual detection.

When the shortcut opens another app

Control, Shift and S is different from Option plus Space. Resolve conflicts rather than repeatedly pressing a competing shortcut.

When words are ready but not inserted

Check Accessibility for the exact installed app and keep the destination unchanged.

When Email formatting fails

Optional AI failure should leave the transcript available; local Email layout can still preserve your words.

When the local writing server is unavailable

Ollama is needed only for the selected local writing path, not plain Dictation.

Check missing first or last words

Wait for Listening, then stop only after the final word. Compare a whole recording, not just a live partial.

When an old trial appears exhausted

The new policy is ten total uses, not ten additional uses on every update.

Sayloom and VoiceInk: an evidence-based comparison

Sayloom does not currently implement every VoiceInk model family or streaming workflow.

Compare proposed price and actual store price

Sayloom’s USD 4.99 is proposed. VoiceInk’s official Mac prices were USD 25, 39 and 49 on this review date.

Compare Mac, iPhone, iPad and Android availability

A similarly named Android app is not proof of an official VoiceInk Android release.

Compare local speech model support

VoiceInk supports additional model families; Sayloom currently ships a Whisper-based local path and Apple Speech option.

Compare email formatting without losing meaning

A fluent result that changes the action fails this task.

Compare exact replacements and vocabulary hints

Dictionary overlap does not mean the two products use the same correction-learning implementation.

Compare privacy by the actual processing path

Both products can have optional providers. Local defaults do not describe every enabled integration.

Compare shortcut ergonomics on your Mac

A shortcut that conflicts on one machine is not a universal product benchmark.

Design a small recognition comparison

Use the same recording and settings. A single sample cannot establish a broad accuracy claim.

Compare Sayloom Desktop with Sayloom Store

The shared bundle identifier does not make the editions functionally identical.