Audio File to Text — Upload Any Audio, Get a Transcript

Not all audio lives on YouTube. Podcast episodes, recorded meetings, lecture recordings, interview files, local video downloads — these don't have a URL to paste. INDXR.AI's Audio Upload tab accepts any audio or video file and runs it through the same AI transcription pipeline used for YouTube videos.

The result is a full text transcript, stored in your library, exportable in every format INDXR.AI supports.

Supported Formats and Limits

Maximum file size: 500MB. Most web upload tools cap at 50–100MB. INDXR.AI's upload pipeline handles larger files without that constraint — we've tested uploads of over 200MB and 210+ minutes of audio without issues.

If you need to split very large files, FFmpeg handles this cleanly: ffmpeg -i large_file.mp3 -t 3600 part1.mp3 -ss 3600 part2.mp3 splits a file at the 1-hour mark.

How the Transcription Works

Once uploaded, the file is processed through AssemblyAI Universal-3 Pro — the same model used when INDXR.AI transcribes YouTube videos without captions. The model handles a wide range of audio conditions:

Languages: 99+ languages with automatic detection. You don't need to specify the language; the model identifies it from the audio.

Audio quality: The model is trained on real-world audio including phone recordings, conference recordings with background noise, and studio-quality content. YouTube's auto-captions achieve 60–80% accuracy on challenging audio — AssemblyAI Universal-3 Pro consistently reaches 94–96%+ on clean recordings.

Punctuation: Unlike auto-captions, AssemblyAI output includes proper punctuation and sentence boundaries. This matters for readability, for SRT/VTT timing quality, and for downstream uses like RAG export where sentence detection affects chunk quality.

Processing time: Approximately 1 minute of processing time per 10 minutes of audio for most files. You receive a live elapsed timer during processing, and the job continues on the server even if your connection drops briefly.

Credit Cost

Audio Upload transcription costs 1 credit per minute of audio, with a minimum of 1 credit. The credit is charged based on the actual audio duration detected after upload, not the file size.

What You Get After Transcription

The transcript appears in your library alongside any YouTube transcripts you've extracted. From there:

Export in any format: TXT plain, TXT with timestamps, Markdown with YAML frontmatter, SRT, VTT, CSV, JSON, or RAG-optimized JSON. SRT and VTT output is resegmented to 3–7 second blocks at 42 characters per line — the professional subtitle standard.

Edit in the rich-text editor: Correct errors, add formatting, and annotate the transcript. Edits are saved separately from the original, so you can always revert.

Generate an AI summary: An AI-generated summary with key points and action items is available for any transcript in your library, at 3 credits.

Export as RAG JSON: For podcasts, lectures, or interviews you want to make searchable via a vector database — enable RAG JSON export to get chunked, metadata-rich output ready for LangChain, LlamaIndex, or direct vector database ingestion. See YouTube Transcripts for RAG Pipelines for the full pipeline.

Common Use Cases

Podcast transcription: Upload an episode MP3 directly. Export as Markdown for show notes, TXT for a newsletter, or RAG JSON to build a searchable podcast archive.

Recorded lecture notes: A 90-minute lecture recording becomes a searchable, editable transcript. Export as Markdown for an Obsidian note, or as CSV for analysis alongside other course materials.

Interview transcription: Upload a recorded interview. The transcript is accurate enough to quote from directly — useful for journalists, researchers, and user researchers.

Downloaded YouTube videos: If you've downloaded a YouTube video that has no captions, extract the audio and upload it to get a transcript the same way as a live YouTube URL.

For YouTube videos (where you have a URL rather than a downloaded file), use the YouTube Transcript Generator instead — it handles auto-caption extraction for free and AI transcription when captions aren't available. For credit packages, see the pricing page.