Casi todo el mundo tiene la intención de preguntarles a sus abuelos sobre su infancia y a la mayoría de las personas se les acaba el tiempo. Grabar la conversación es la parte fácil ahora. Esta guía cubre cómo realizar la entrevista, convertir el audio en texto con capacidad de búsqueda en su propio dispositivo y mantener un archivo familiar que sus nietos podrán abrir.
Almost everyone means to do this. The thought arrives at a birthday, or in a hospital corridor, and you promise yourself that next visit you will bring something to record with and ask your grandmother about the crossing, or the winter she has mentioned twice and never explained.
Then the visits go by. There is a cake to cut and a cousin to catch up with, and it never feels like the moment to put a phone on the table.
What you are left with is three anecdotes worn smooth from retelling and a box of photographs nobody alive can name. The details that made the stories real, the name of the street, why she never spoke to her brother again, are simply gone.
The recording part is easy now. Every phone in the house is a competent audio recorder and it costs nothing. The hard part has moved somewhere less obvious, and it is why so many good recordings end up doing nothing.
You come home from a good afternoon with three hours of audio and enormous relief, because the thing you kept postponing is finally done. You put the file somewhere safe. Then you never open it again, because opening it costs three hours you do not have.
Six years later somebody asks about the boat. You know the story is in there. You do not know whether it sits at eleven minutes or at two hours and seven, and nobody scrubs three hours of audio hunting for one story, so it stays in the file: technically preserved, practically unreachable.
Audio is opaque. You cannot skim it, search it, or paste a paragraph into a message to your sister. Text does all of that, and the conversion between the two is the whole difference between a recording and an archive.
Without Yaps
Four long recordings sit in a folder called Grandma with filenames like VOICE_0142. Nobody knows which one has the story about the boat, or which voice is the aunt who visited that day. Playing them means committing an evening, so nobody does, and the collection quietly becomes a thing the family feels guilty about.
With Yaps
The same four recordings become transcripts on your own computer, labelled with who was speaking, filed in a private vault you can search. Type boat and you land on the paragraph. You can read it, quote it, export it as plain text, and hand the whole thing to a cousin who will actually open it.
Think about what is actually said in a good family interview, once the person relaxes and stops performing.
Medical history, and what people died of. Money, including who lent whom what and who never paid it back. How a border was crossed. A sibling's drinking. A first marriage the grandchildren were never told about. A wartime memory somebody carried for seventy years and never said out loud until you sat down and asked properly.
That is not a podcast. It is the most sensitive audio your family will ever produce, and it is about people who cannot consent for themselves once they are gone.
So what happens when you upload it? Most transcription tools send your file to a company, where it is processed and stored under terms you did not write, for a period you did not choose. Your grandmother agreed to tell you about her mother. She did not agree to become a row in somebody's storage bucket.
Doing the transcription on your own computer removes the question. The audio stays on hardware you own, the text lands on the same machine, and the only people who hear the recording are the ones you play it to.
Yaps is a voice assistant that runs on your own device. For an oral history project it does four useful things, all of them locally.
It turns the recording into text, offline. The Yaps Studio editor imports an audio file you have and transcribes it on your machine into text or an SRT subtitle file. Nothing is uploaded. A long interview becomes a document you can read in twenty minutes instead of an afternoon you keep postponing. If you filmed the conversation instead, turning video into a transcript offline covers that path.
It can label who said what, on Windows and macOS. Interviews are where this matters most, because a transcript that runs the questioner and the answerer together is hard to read. In desktop Studio on Windows and macOS you install a small speaker identification pack, about 32 MB, from the Features screen, then switch the toggle on before you transcribe. It is optional, off by default, runs on your device, and is not available on Android or Linux. It does not mean Yaps sits in on live calls.
It files the transcript in a private, searchable vault. Transcripts land in a local vault alongside your own notes, searchable without any cloud index of your family's business. Voice notes support plain text, checklists, and kanban boards, so a project with six relatives can be a board with a card each, and everything exports to Markdown and plain text.
It answers questions across the whole collection, on your device. Once several interviews are transcribed, you can ask across your own notes and get an answer, and that runs on a local model rather than a server. The piece on using an AI memory app goes deeper on how that works.
One more thing worth knowing. Yaps reads text back aloud, a pleasant way to review a long transcript while you cook, with 18 voices on desktop and 2 on mobile. Those voices are English speakers in practice, so if your grandmother told her stories in Tagalog or Twi, the transcript will be in her language but the read-aloud will not suit it, and a multilingual cloud reader is the honest recommendation. See text to speech.
That one question sorts the options faster than any feature list. Hold the on-device approach against what families usually do instead.
| What matters | Yaps (record yourself, transcribe offline) | Cloud transcription service | Leave it as audio |
|---|---|---|---|
| Recording leaves the family | No, it stays on your machine | Yes, the file is uploaded | No |
| You can find one story in seconds | Yes, local search | Yes, in the vendor account | No |
| Works with no internet | Yes | No | Yes, but unsearchable |
| Labels who said what | Optional, Windows and macOS desktop | Commonly yes | No |
| Cost as the collection grows | Flat subscription, hours do not add up | Commonly charged per minute or hour | Nothing, and worth nothing |
| Ask questions across every interview | Yes, on your device | Varies, in the cloud | No |
| Repairs old damaged tape audio | No | Some offer clean-up | No |
Read the last row as a real limit. Yaps does not restore or enhance damaged audio, so a warped cassette from 1978 needs dedicated restoration software first. Everything above that row is where keeping it local earns its place: a tool that never receives your grandmother's recording cannot lose it, cannot retain it after you delete it, and cannot change its terms next spring.
The cost row deserves the plain version too. On Android there is a free tier, so the phone side costs nothing to try. On desktop, meaning Windows, macOS, and Linux, there is no free tier: you get a 7-day free trial and then pick a plan, Pro at $15 a month or Max at $25 a month. Since the transcribing happens on the desktop side, that flat plan is the number that matters as the recorded hours pile up.
Your grandmother agreed to tell you her stories. She did not agree to have them processed by a company neither of you has heard of.
Yaps
The whole loop costs about two hours for a two hour interview, mostly the reading pass in step five.
Say what you want to record and why, and promise to stop or delete anything they would rather not keep.
Use the phone's own recorder app. One microphone, close to the speaker, hard table, no television in the background.
Keep the original untouched as the master and work from a copy. This is the file everything else depends on.
Import the audio and let it become text on your own machine. Switch on speaker identification first if you are on Windows or macOS.
Fix place names and family spellings, then record the date, the place, who was there, and who was asking.
Into the searchable vault, exported to plain text, and copied somewhere that is not your house.
The interview matters far more than the equipment, and most people get it wrong the same way.
Do not ask for a chronology. Starting at the beginning produces dates and school names in the flat voice people use for facts. Nobody listens to that recording twice.
Ask about objects instead. Bring a photograph, a tool from the shed, a recipe card in her mother's handwriting, a chipped bowl. Put it on the table and ask what it is. Objects unlock the sensory in a way that questions about decades never do. Instead of asking about her childhood, ask what the kitchen smelled like on a Sunday.
Ask about the ordinary. How much the rent took out of a wage. What you did when somebody got sick and there was no money. Who did the washing. These vanish first, because they were too obvious to mention, and they are what makes a life legible to a grandchild.
Use one microphone and put it close. A phone flat on a hard table half a metre from the speaker beats an expensive setup across a soft living room. Turn off the extractor fan, the radio, and the television. Background noise is the biggest cause of a messy transcript.
Let silence run, and do not correct anybody. Count to five after an answer seems to have finished, because the most valuable sentence often arrives in second four. If a date is wrong, let it stand and check later. Corrections turn a conversation into an examination.
Plan two sessions, not one. Ninety minutes is a lot for an older person, and the second conversation is almost always better, because they have spent the week remembering what they forgot.
Afterwards, sit in the car and speak your own notes while the afternoon is fresh: what she avoided, what to follow up next time. That habit fits into a journaling practice, and the best AI journal app guide covers the setup.
An archive is only an archive if it survives the technology it was made on. This part is unglamorous, and it fails most often.
Keep boring formats. Plain text and Markdown transcripts, and uncompressed WAV audio, will still open in fifty years because they are simple and universal. Yaps exports notes to Markdown and plain text, and Studio exports audio as WAV and subtitles as SRT. Avoid anything that opens only inside one company's app, including this one.
Keep three copies in two places. One on your computer, one on an external drive, one somewhere that is not your house. A fire takes everything in a single room, and family recordings tend to live in one room.
Name files so a stranger understands them. A date, a surname, and a session number is enough. The pattern 1953-margaret-okafor-session-02 tells somebody in 2070 what they have. VOICE_0142 tells them nothing.
Write one index note. A single document listing every interview, who it is with, when it happened, and two lines on what is in it. Without that map, a large collection becomes another kind of shoebox.
Share it while people are alive. Send the transcripts to siblings and cousins now, because relatives correct details and add stories once they read what somebody else remembered.
An honest scope is more useful than a feature list, and here the boundaries matter.
It does not record the conversation for you. There is no meeting bot, no automatic capture, and no live call transcription. You record with your phone's own recorder app, then import that file into Studio. That boundary is deliberate.
Speaker labelling is Windows and macOS desktop only. It is optional, off by default, and needs a small pack of about 32 MB installed from the Features screen with the toggle switched on before you transcribe. It is not available on Android or Linux, so on those platforms label the speakers yourself during the correction pass.
It does not repair old or damaged audio. Hiss, tape wobble, a recording made across a noisy room in 1978: Yaps transcribes what is there and improves nothing. Restore the audio with dedicated software first, then transcribe the cleaned file.
It does not do genealogy. No family tree, no records search, no census matching, no DNA. Yaps handles the spoken part of family history and nothing else.
There are no integrations. Yaps does not connect to family tree platforms or archive systems. It produces clean, structured, paste-ready text you copy wherever you need it, plus Markdown and plain text exports.
There is no iPhone or iPad app yet. Yaps runs on Android, Windows, macOS, and Linux today, plus a Chrome extension, and iOS is coming soon. You can still record on an iPhone and transcribe the file on a computer.
Read-aloud voices are English speakers in practice. Dictation handles about 25 spoken languages and detects them automatically, so transcripts are fine in other languages. The voices are the gap, and a multilingual cloud reader suits that better.
This is where the work pays off, and it is the part people do not anticipate.
Once four or five interviews sit in your vault, you stop having a set of documents and start having something closer to a family memory. You can ask a question across all of it, and the answer comes back from a model running on your own machine, offline.
Where did Grandad say his father worked before the factory. Who else mentioned the house on Cardigan Street. What did three people say about the same funeral. No filename structure answers those, and they are the questions that surface when somebody is writing a speech or settling an argument.
Treat the answers as a way of finding the right passage rather than the truth. Open the transcript it points to and read the actual words, because their phrasing was always the point.
Start with Yaps, and more importantly, start this month. The tool matters far less than the calendar, because the constraint here has never been software. It is that people die on their own schedule and nobody warns you which visit was the last one.
The workflow is small enough for a Sunday. Ask properly and get a yes. Record on the phone already in your pocket. Import the file into Studio and let it become text on your own computer, with speaker labelling on if you are on Windows or macOS. Correct the names, file it in a searchable vault, and back it up off site. What you get is what the family actually wanted: a document anybody can read, search, and quote.
Two honest exceptions. If your recordings are old and damaged, restore the audio with dedicated software first. If you need read-aloud in another language, or you work only on an iPhone, use a multilingual cloud reader or wait for the iOS release.
Otherwise the next step is a phone call to arrange a visit. Bring a photograph and a question about a chipped bowl.
Yaps is the strongest default for the part that usually fails, which is turning the recording into something usable. You record with your phone's own recorder, then import that file into the Yaps Studio editor, which transcribes it offline on your computer into searchable text. On Windows and macOS you can optionally switch on speaker identification so the transcript shows who said what.
Ask permission, then record with the phone recorder app you already have, placed on a hard table about half a metre from the person speaking. Turn off the television and the extractor fan, ask about objects rather than dates, and let silences run for a count of five. Plan two sessions of roughly ninety minutes rather than one long one, then transcribe the audio offline in Yaps Studio.
Yes. Yaps Studio transcribes imported audio files entirely on your own device, so the recording never leaves the computer it is on and no internet connection is needed. That matters here because family recordings hold health details, money, and private history about people who cannot consent for themselves once they are gone.
No. Yaps has no meeting bot, does not join calls, and does not record live conversations automatically. You capture the interview with your phone's own voice recorder, then import that audio file into Studio to be transcribed offline.
Yes, on Windows and macOS desktop only. Speaker identification is optional in desktop Studio: install a small pack of about 32 MB from the Features screen, then switch the toggle on before transcribing, and it runs on your device. It is off by default and not available on Android or Linux, so on those platforms you label the speakers yourself while correcting the transcript.
Ask about objects and about ordinary daily life rather than asking for a chronology. Put a photograph, a tool, or a recipe card on the table and ask what it is. Ask what the kitchen smelled like on a Sunday, what the rent took out of a wage, or who did the washing. Objects and routines produce the detail that makes a life legible to somebody who never knew it.
Yes, and it is better treated as a courtesy than a legal formality. Tell the person you would like to record their stories, explain that the audio stays on your own computer, and offer to stop or delete anything at any point. Recording rules also vary by country and by state, so check what applies where you live.
Transcribe it, then search the text. Audio cannot be skimmed, which is why long recordings go unopened for years, but a transcript in a searchable vault turns the story about the boat into a two second search. Once several interviews are transcribed, Yaps can also answer questions across all your notes at once, on a local model on your own device.
No. Yaps does not restore, denoise, or enhance audio in any way. It transcribes what is on the file, so hiss, tape wobble, and a conversation recorded across a noisy room will produce a rough transcript. Run damaged material through dedicated audio restoration software first, then transcribe the cleaned version.
No. There is no family tree, no records or census search, no DNA matching, and no integration with any genealogy platform. Yaps handles the spoken side of family history: turning interviews you recorded into private, searchable text. Use a dedicated genealogy service for the documentary research.
On Android there is a free tier, so you can start using Yaps on your phone at no cost. On desktop, meaning Windows, macOS, and Linux, there is no free tier: you get a 7-day free trial and then choose a paid plan, Pro at $15 a month or Max at $25 a month. Starting that desktop trial requires an account, which covers access and billing, not the transcribing: that still happens on your own computer.
Not yet. Yaps runs on Android, Windows, macOS, and Linux today, plus a Chrome extension, and the iPhone and iPad app is coming soon. For this workflow that matters less than it sounds, because you can record on an iPhone and transcribe the file in Studio on a computer.
Yes, for the text. Yaps dictation recognises about 25 spoken languages and detects them automatically, so interviews in many languages are covered for the transcript side. The honest gap is the read-aloud voices, which are English speakers in practice, so a multilingual cloud reader fits better if you want a transcript read back in Tagalog or Polish.