Skip to content
INDXR.AI

Workflows · 7 min read

Audio to text: interviews, lectures and podcasts you can search and quote

vintage microphone head resting on sandy ground casting long shadow
IE
INDXR.AI Editorial
Published April 16, 2026 · Updated November 9, 2026

You recorded an interview, sat through a lecture, or have a podcast episode you need to work from. Upload the file and INDXR gives you the full text back: punctuated, split by speaker, with a timestamp on every paragraph. The transcript stays in your library with its source, so you can search it later, check a quote against the exact moment in the recording, and ask Claude or ChatGPT about it.

It costs 1 credit per minute of audio. A free account includes 10 credits, enough for 10 minutes, so you can run a real recording through it before you pay anything. There is no subscription, and credits never expire.

Who it is for

Interviews. A journalist, a researcher or a writer with a stack of recorded conversations. You rename the speakers once, find every place someone said a thing, and check the wording against the recording before you quote it. There is a full walkthrough in transcribing interviews.

Lectures. A student with recorded classes, or a course you downloaded. Put the lectures in one collection and search the whole course for a term, or have an assistant write study notes with a reference per section.

Podcasts. An episode you want to cite, or your own show you want in writing. Long files are fine: up to 10 hours each.

How it works

You upload the file. Uploading needs a free account, no card required. Audio and video both work, up to 500MB and 10 hours per file. You do not pick a language: the model detects it from the audio.

It transcribes while you wait. An hour of audio is usually ready within a few minutes, two hours within about a quarter of an hour. It runs on the server, so closing the tab does not cost you the job. Everything is processed on European infrastructure, and the uploaded file is deleted once transcription finishes.

It lands in your library. There you can correct it, search it, put it in a collection, summarise it, export it in seven formats, or let Claude or ChatGPT read it.

INDXR's cost panel after a file is added: the recording's title, its length of 54:56, an AI transcription label, and a total of 55 credits shown before you start, with the balance left afterwards.
After you add a file, INDXR reads its length and shows the exact cost before anything starts: one credit per minute, here 55 for a recording just under an hour.

What the transcript looks like

Transcripts come back punctuated, with sentence boundaries the model found, rather than one long block of lowercase text. Speakers are detected and labelled, and you can rename them: change Speaker A to the interviewer's name once and every occurrence updates, in the transcript and in every export, with the original label always recoverable. Text is grouped into paragraphs, each with a timestamp for where that passage begins.

A finished transcript where each paragraph opens with a timestamp and a bold speaker name, here Sarah Chen and Dr. Miguel Ferro, followed by their words.
A finished transcript reads back with punctuation, a timestamp per paragraph, and speaker labels you rename once to update everywhere.

Check every quote against the recording

A transcript is a strong draft of what was said, not proof of it. Before a sentence goes into a paper or an article, it is worth hearing it once. Because each paragraph carries the moment it starts, that takes seconds: search the library for the phrase, open the transcript at that paragraph, and play it back from there.

Take a two-hour interview transcribed in March. That month you write your piece from it. In June a quote is questioned, so you search for the phrase, jump to the timestamp and check what was actually said. Nothing was re-uploaded, and none of it cost extra.

The assistant searches your transcripts and answers with the passage and the minute it was said, so you can check its answer the same way you check your own. More on this in using Claude or ChatGPT with your own sources.

How accurate it is

English transcription accuracy here is close to the current technical ceiling. AssemblyAI places English in its highest accuracy band, below ten per cent word error rate, and on independent benchmarks Universal-3.5 Pro posts an English word error rate in the low single digits.

Accuracy falls where you would expect: strong accents, overlapping speakers, background noise, technical vocabulary, unfamiliar names. On difficult audio, expect to correct a few names. For languages other than English, AssemblyAI publishes a per-language accuracy table worth checking before you commit a long recording.

What you can upload

Audio or video, in fifteen formats. Upload the file as it reached you: there is no conversion step and no question about which container you have.

Audio formatsMP3, MPGA, M4A, AAC, WAV, OGG, OPUS, FLAC
Video formatsMP4, MPEG, WEBM, MOV, FLV, AVI, MKV
Maximum file size500MB
Maximum length10 hours per file
Languages99, detected automatically

Where the common ones come from, and what to know about each:

  • M4A is what the iPhone Voice Memos app records. Upload it as it is; converting it to MP3 first only adds a step and loses a little quality.
  • OPUS is what WhatsApp exports a voice message as. OGG is the open container Opus audio usually travels in. Both upload without renaming.
  • WAV is uncompressed, which is why a recorder set to archival quality writes it and why the files are big: an hour of stereo WAV at CD quality is roughly 600MB, over the 500MB limit. Split it at a pause rather than compressing it.
  • FLAC is lossless. It does not transcribe better than a good MP3 of the same recording, so upload it because it is the file you have, not for accuracy.
  • MOV is what an iPhone camera films; MKV is what many downloads and archives arrive as; WEBM is what a browser writes when it records your screen. All upload directly.

The full upload flow and its checks are in the uploads guide.

Video files

A video is transcribed the same way as audio: the audio track is taken out and the picture is discarded. That has two consequences worth knowing. The transcript does not describe what happens on screen, so a screen recording gives you what was said about the screen. And text that appears in the picture, such as slides or burned-in captions, is not read.

Here is a measured one. A clip of just over five minutes from a podcast interview, uploaded as a 7.4MB MP4. It cost six credits, came back with two speakers labelled, split into 74 timed passages, ready in under half a minute, and the model reported 96.5 per cent confidence.

A finished transcript of the video: each paragraph opens with a timestamp and a speaker label, Speaker A and Speaker B, followed by their words.
The transcript of a video file reads back punctuated, split into paragraphs with a timestamp each, and labelled by speaker.

The same transcript exports as SRT or VTT subtitles, rebuilt to broadcast conventions: at most 42 characters per line and 2 lines per cue, breaking on sentence boundaries where it can:

2
00:00:02,920 --> 00:00:07,338
Look, you get out in that world, Joe,
you're famous everywhere.

If subtitles are what you are after, the SRT generator article covers them in full.

Common questions about uploading

Which audio and video formats can I upload?

MP3, MP4, MPEG, MPGA, M4A, AAC, WAV, WEBM, OGG, OPUS, FLAC, MOV, FLV, AVI, and MKV, up to 500MB per file. You do not need to convert anything first.

Can I transcribe a voice memo from my iPhone?

Yes. iPhone voice memos are M4A files, which upload directly. The limits are the same as for any file: up to 500MB and 10 hours.

What happens if my file has no speech in parts of it?

Silence and background noise are skipped in the transcript. You are charged per minute of audio, rounded up, so a recording with long pauses costs the same as one without.

Do I need an account to upload a file?

Yes, uploads require an account, because transcription runs on credits and the transcript is saved to your library. New accounts start with 10 credits, so you can transcribe a short recording before deciding whether to buy more.

What else you can do with the transcript

You can correct the text without losing the original: edits are stored separately and the untouched version stays one click away.

You can have the recording summarised into chapter notes, with a timestamp per chapter that takes you back to that moment.

The library list: rows of saved transcripts, each with its title, how it was made, its duration and date, and all of them openable again.
Every transcript stays in your library. A free converter hands you one download; here you keep an archive you can reopen, search and re-export.

And you can export in seven formats. Everything is free except the RAG export, which chunks the transcript with metadata for vector databases.

Scroll the table horizontally →

GroupDownloadContentsCost
TextTXT — plain textNo timestampsFree
TextTXT — with timestamps[HH:MM:SS] per lineFree
TextMarkdownNotion, Obsidian, blogFree
TextMarkdown — with timestampsSections per timestampFree
SubtitlesSRTSubRip SubtitleFree
SubtitlesVTTWeb Video TextFree
DataCSVSpreadsheet compatibleFree
DataJSONsegments with start/end timeFree
DeveloperRAG JSONLangChain, LlamaIndex, Pinecone1 credit / 10 min

What it costs

1 credit per minute of audio, rounded up, based on the duration detected after upload rather than the file size. Prices below are on the Plus package, €25 for 1,000 credits.

Audio lengthCreditsCost at Plus pricing
1 minute1 credit€0.03
10 minutes10 credits€0.25
30 minutes30 credits€0.75
1 hour60 credits€1.50
2 hours120 credits€3.00

There is no subscription, and credits never expire. Most transcription services sell a monthly plan instead. One 2024 survey found that six in ten people had avoided subscribing to a service because they expected cancelling to be difficult. Here there is nothing to cancel. If you are weighing this against a subscription tool, see INDXR vs Otter.ai.

Why not use a free converter

For one short recording whose wording does not have to be exact, a free converter is fine. What you trade, when the result does matter, is quality, headroom and the archive. A free service has to keep its own costs down, which generally means a lighter model and tighter limits, and it hands you one download and forgets it.

Scroll the table horizontally →

A free converterINDXR
OutputUsually raw words: little punctuation, no paragraphs, no speaker labelsPunctuated, split into paragraphs, speaker-labelled and timestamped
After you downloadOne file, then it is goneKept in your library with its source, searchable, and readable by Claude or ChatGPT
Export formatsUsually one, plain textseven: TXT, Markdown, SRT, VTT, CSV, JSON, and RAG JSON
File size limitOften 50 to 100MB500MB
Where it is processedUsually unstatedThe EU, and never used to train a model

Try it on your own recording

The only reliable way to judge a transcription service is to run something through it that you care about. A free account includes 10 credits, covering 10 minutes of audio, with no card. If the result is good, the credits you buy afterwards do not expire, and every transcript you make stays in your library.

Frequently asked questions

Is my recording used to train AI models?
No. The file you upload is deleted on our side as soon as the transcription finishes, and only the transcript text stays in your library. Our transcription provider, AssemblyAI, does not use your audio to train its models, because we have opted out of its model-improvement programme, and its own data retention is set to one day, the shortest period it offers. Everything is processed inside the EU.
Can I convert an MP4 or a MOV to text?
Yes. INDXR takes the audio track out of the video for you, so you upload the file as it is. MKV, MOV, AVI and the other video formats work the same way, up to 500MB per file.
What happens if the transcription fails?
Your credits are refunded in full. The cost is reserved from your balance when the job starts, and it is booked straight back on any failure, including when the server goes down in the middle of the job, so a transcription that does not complete never costs you anything.
What if I run out of credits partway through a file?
It cannot happen. The full cost of the file is reserved before transcription begins, based on the duration we detect, so a job you cannot fully afford is declined up front rather than started and stranded halfway.
Can I upload a recording of a meeting or a phone call?
Technically yes, any audio file works. Bear in mind that recording a conversation legally requires the consent of the people taking part in many countries, and making sure you have that consent is your responsibility, not something the tool can check for you.
Why are names and technical terms transcribed wrong, and what can I do about it?
Proper names and specialist jargon are the known weak spot of every speech model, because it works from sound and cannot guess a spelling it has never met. Correct it once in the editor: every export you make afterwards carries the correction, and when there are several speakers you rename a speaker label once and it updates everywhere it appears.
Does it work when people talk over each other?
Partly. The model detects who is speaking and labels each speaker, but where voices overlap the boundaries get less precise and a word can end up attached to the wrong speaker. On clean turn-taking the labelling is reliable; through a stretch of crosstalk, expect to fix a few lines.
How do I split a file that is too large or too long?
Cut it at a pause between sentences rather than in the middle of one, because a word split across the seam can be lost. Any everyday audio editor can do this. Sending the parts separately costs nothing extra, because you pay per minute of audio, not per file. The limits are 500MB and 10 hours per file.

Sources

See also