Skip to content
SoloWOS
Start your free trial

AI for freelancers

From voice memo to tasks, hours and expenses

By SoloWOS teamPublished

IN SHORT

To turn a voice memo into tasks, say each action with its numbers: the hours and the project, the amount and what you bought, and what you'll send to whom. A transcription service turns the audio into text, and an assistant splits the text into entries. Check names and amounts before relying on them, because speech-to-text mishears both.

Prices and limits as of .

Key takeaways

  • One sentence per action works best: hours on a project, an amount spent and what for, and a document to prepare for a client.
  • Speech-to-text mishears names and numbers more than anything else, so say the client's name as it's saved and say amounts with the currency.
  • In SoloWOS, a voice note to the assistant can be up to 2 minutes. It uses transcription minutes (5 a month on Basic, 300 on Solo) and one assistant request (10 on Basic, 500 on Solo).
  • The assistant creates the time entry and the expense right away, because they don't reach anyone. A quote it drafts stays a draft, and sending it waits for your Confirm.
  • A voice note doesn't attach a receipt. The IRS expects records that show the payee, amount, date and what you bought, so add the receipt to the expense afterward.

Voice-to-task apps are good at one thing: turning a rambling memo into a to-do list. A freelancer walking out of a client meeting needs something else. The three hours go in the time tracker, the parking goes in expenses, and "I'll send her the quote tomorrow" should become a quote draft, not a reminder to write one. That only works when the tool that hears you can also write to the places where your hours and expenses live.

How a voice note becomes three entries

There are two steps. First, a speech-to-text service turns the audio into text. Services like Deepgram and OpenAI's transcription API return the words with punctuation, but they can't know that "Ana" is a client in your records or that "forty" was an amount. Second, an assistant reads that text, splits it into separate actions and matches each one to your data: the client, the project, the expense category. The second step is where mistakes get caught, if the assistant asks before guessing.

How to say it so it comes out right

  • One action per sentence. "I worked three hours on Ana's logo. I spent forty-five dollars on an Uber to her office. Tomorrow I'll send her the quote for the brand guide."
  • Use the client or project name as it's saved in your records. If the project is called "Café Luna rebrand", say that, not "the coffee people".
  • Say amounts with the currency: "forty-five dollars", not "forty-five". Say decimals as "twelve fifty" only if the context makes it clear.
  • Say the date if it isn't today: "Yesterday I worked two hours on…"
  • Say whether time is billable if it isn't obvious: "one hour on the proposal, not billable".
  • Keep it under a minute. A long, wandering note produces a long list of things to check.

WORKED EXAMPLE

A 20-second voice note after a client meeting (illustrative)

Marco records: "Today I worked three hours on Ana's logo. I spent forty-five dollars on an Uber to her office. Tomorrow I'll send her the quote for the brand guide." The text appears in the box and is sent. The assistant finds Ana's only project and logs a 3-hour billable time entry on it, ending now. It saves a $45 expense dated today in Transportation and fuel with the vendor Uber. It creates a quote draft for Ana titled "Brand guide" with an empty price, because Marco didn't say one, and replies with links to all three. Nothing reached Ana. The note used 1 transcription minute (20 seconds, rounded up) and 1 assistant request.

In SoloWOS: the microphone in the assistant

  1. Open the assistant with Ask AI at the top of any page (or Ctrl/⌘ + J), or use the "What do you need today?" box on Home. Both have a microphone button.
  2. Tap it and speak. A counter shows the time; notes stop at 2 minutes. Tap the red square to stop, or Cancel to throw the note away.
  3. What was understood appears in the box and is sent after 3 seconds ("Sending what you said in 3 s."). Tap Edit first, type, or tap the box to stop the countdown and fix a name or an amount before sending.
  4. The assistant splits the note into actions. Hours go to a time entry for you (never a running timer), money spent goes to an expense, and "I'll send the quote" becomes a quote draft for you to review.
  5. If a client or amount doesn't match your SoloWOS data, or a client has several projects, it asks instead of guessing. Anything that would reach a client, like sending the quote, shows as a card with Confirm and send.

The audio goes to SoloWOS's transcription provider (Deepgram or OpenAI) only to turn it into text, with your app language as a hint, and isn't stored by SoloWOS, as a file or anywhere else. The text then goes to Anthropic like any request you type; under Anthropic's commercial terms it isn't used to train their models. Minutes are counted per note, rounded up, and only when the transcription worked (a note with no words still counts, because the provider was paid, and you're told). You can record up to 20 notes an hour. The button only appears when the browser can record and the page is on HTTPS, and hours need Time Tracking to be on.

What to check after a voice note

  • The amount and currency on the expense, and the category the assistant picked.
  • The project and date on the time entry, and whether it's marked billable.
  • The client on the quote draft, and every line before you send it. Prices the note didn't mention stay empty.
  • The receipt. A voice note creates the expense without one. Open it and attach the photo, or use Read receipt in Expenses next time for anything with a paper slip.

That last point matters at tax time. The IRS says supporting documents for an expense should show the payee, the amount, proof of payment, the date and a description of what you bought, and that proof of payment alone doesn't establish a deduction. An expense line you dictated is a reminder, not the record. Log business expenses with receipts so each one has its slip attached.

When typing is faster

Voice wins when your hands are busy or you're between places. At a desk, typing "3h Ana logo, $45 Uber" into the assistant takes a few seconds, doesn't use transcription minutes and can't mishear anything. Many freelancers use both: voice on the way out of a meeting, typing for the rest. To see what drafting and logging by AI saves you in a month, try the AI time savings calculator with your own numbers. Everything here runs through the SoloWOS AI assistant; more posts are in AI for freelancers.

Questions

Can I log business expenses by voice?

Yes, if the tool that hears you can write to your expense log. In SoloWOS, saying "I spent forty-five dollars on an Uber to a client" in a voice note creates the expense with the amount, date and a category to check. Attach the receipt afterward, since a dictated line isn't a receipt.

What if it mishears a client's name or an amount?

The text appears in the box for 3 seconds before it's sent, so you can tap Edit first and fix it. If it's sent with a name or amount that doesn't match your SoloWOS data, the assistant is told to ask you rather than guess. You can also edit any entry it created.

Is my voice recording saved?

No. SoloWOS sends the audio to its transcription provider to turn it into text and doesn't store it, as a file or in the database. Only the text continues to the assistant, and SoloWOS keeps the minutes used to apply your monthly limit.

Does it understand Spanish?

Yes. The transcription uses your app language as a hint, and the assistant answers in the language you use. If you mix languages in one note, check names and amounts a little more carefully.

Sources

Facts checked against these sources on October 24, 2026. Rules and amounts change, so confirm them at the source before you file or sign.

We don't get paid by any brand on this page: no affiliate links, no sponsored placements.

  1. Deepgram docs: Pre-recorded audio (speech to text)
  2. OpenAI API docs: File transcription (speech to text, 25 MB file limit)
  3. IRS: What kind of records should I keep
  4. Anthropic: Commercial Terms of Service (customer content isn't used to train models)

Last reviewed:

SoloWOS tools mentioned here

Run your client work from one account

14 days free, no card needed · Then Basic for $2/month.

Start your free trial