Audio Summary for Meetings, Podcasts, and Lectures
What an AI audio summary gives you, and what it leaves out
Summarize a recording such as a meeting, interview or lecture with a chosen length and format. Add a focus to emphasize decisions, action items or another subject, then review the result against the recording. A summary condenses the discussion and may omit details. Use Audio Transcription when you need the spoken words, speaker labels or timestamps instead.
Cost: Free within the available allowance. Usage limits apply; see the guide below.
How to summarize an audio file
Upload one recording
Click Upload audio to summarize or drop a file onto the preview area. Audio (MP3, WAV, M4A, AAC, FLAC, OGG, Opus, WebM) and video (MP4, M4V, MOV, MKV, AVI, WebM, MPEG, OGV) up to 200 MB work. A matching file you already uploaded in another tool is listed under Files.
Pick a length
Choose a target in Summary Length: 100, 250, 500, 1000 or 2000 words. 500 words is preselected.
Choose the layout
Select Paragraph or Bullet Points under Summary Format.
Add a focus if you need one
Type the topics that matter into Focus Area (Optional), for example
decisions and action items with owners. Leave it empty for a general summary.Generate, check, download
Click Generate Summary. The summary replaces the player in the preview area. Use Copy to Clipboard, TXT or DOCX, and check names, numbers and decisions against the recording before you share it.
What each setting changes
All three settings become written instructions for the AI model. They steer the result, but nothing checks or trims the text afterwards.
- Summary Length
- 100, 250, 500 (default), 1000 or 2000 words. The number is a target, so expect some deviation. Match it to the recording: as a rule of thumb people speak about 130–160 words per minute, so a 5-minute voice memo holds roughly 700 words and suits 100 words better than 1000.
- Summary Format
- Paragraph is the default and writes the summary as flowing prose, which reads well in reports and emails. Bullet Points asks for a bulleted list, which is easier to scan as meeting notes and pastes cleanly into tickets or chat.
- Focus Area (Optional)
- Your text is added to the instruction as “Focus particularly on: …”, so short noun phrases fit best:
action items with owners and deadlines,definitions and exam topics,pricing objections. The general request to extract key points stays in place, so a focus shifts the emphasis rather than replacing the summary.
Starting points for common recordings
Suggested combinations; there are no preset buttons.
Team meeting or call
250 or 500 words, Bullet Points, focus decisions, action items and owners.
Lecture or training session
1000 words, Paragraph, focus key concepts, definitions and examples, so each topic gets room, not only the main thesis.
Podcast episode or interview
500 words, focus on the guest's main arguments and recommendations. There is no speaker-labelling option, so name the people you care about in the focus.
Voice memo
100 words, Paragraph, no focus. A few minutes of speech rarely contains enough for a longer summary.
What happens to your recording
The upload goes to our server. FFmpeg takes the first audio track, converts it to mono FLAC at 16 kHz and removes metadata; for videos the picture is discarded at this step. That copy is sent to Google's Gemini API (currently the Gemini 3.5 Flash-Lite model) together with the instruction built from your settings. If Gemini is rate-limited at that moment, the same request goes through the OpenRouter service to the same Gemini model.
The temporary FLAC file is deleted from our server as soon as the model has answered, and our server asks Gemini to delete its uploaded copy after the request. Your original upload, the summary text and the DOCX file follow the storage periods listed under Formats and limits.
A summary is written by a language model, not copied word for word. It can drop a point, blend two speakers' views or get a number wrong, especially with crosstalk or a poor microphone. When exact wording matters, run Audio Transcription on the same file; it stays under Files, so you don't need to upload it again.
The TXT download contains the text exactly as shown, including Markdown characters such as ** or # if the model used them. The DOCX file turns those into bold text, headings and lists.
Formats and limits
- Input formats
- MP3, WAV, M4A, AAC, FLAC, OGG, Opus and WebM audio; MP4, M4V, MOV, MKV, AVI, WebM, MPEG/MPG and OGV video
- Audio used
- The first audio track only; the video picture is ignored
- Output
- Text on the page, downloadable as
summary.txtand as a Word file,summary.docx - Maximum file size
- 200 MB
- Files per job
- 1
- Processing
- On our servers with FFmpeg; the summary itself is written by Google's Gemini API
- Watermark
- None
- Usage limits
- Hourly and daily limits apply; some count the number of summaries, others the total recording minutes. An error message tells you when you reach one
- Storage
- Uploads and results are deleted automatically after 48 hours for guests and after 30 days for free accounts. Paid plans can keep files longer.
Frequently asked questions
Is this AI audio summarizer free?
Yes, you can summarize recordings without an account or payment. Because every summary uses an AI model, hourly and daily usage limits apply; they can count the number of summaries as well as the total recording minutes.
Is my audio sent to an AI service?
Yes. A mono 16 kHz copy of the first audio track, plus your focus text, goes to Google's Gemini API, or through OpenRouter to the same Gemini model when Gemini is rate-limited. The picture of a video is never sent.
Why doesn't the summary match the length I chose?
Summary Length is passed to the model as a target and the result is not cut or padded afterwards. A target can also exceed what was said: a 10-minute recording holds roughly 1,300–1,600 spoken words, so 2000 words asks for more text than the recording contains and invites repetition.
What language will the summary be in?
There is no output-language setting. If you need the text in a specific language, Audio Transcription has a language setting and optional translation.
Can I summarize a video or screen recording?
Yes, only the sound is used. The tool reads the first audio track, so if your recording software stores the microphone on a second track, that voice is not included in the summary.
My recording is larger than 200 MB. What can I do?
Uncompressed CD-quality WAV (44.1 kHz, 16-bit stereo) needs about 606 MB per hour and hits the limit after roughly 20 minutes. Converting it to MP3 at 64 kbps with Audio Converter needs about 27.5 MB per hour. Prior lossy compression can affect recognition, so check that speech remains clear in the smaller recording. If one summary is too coarse for a multi-hour session, cut it with Split Audio and summarize each part.