Gemini API Cost

The short answer: roughly $0.01 per one-hour meeting with the default model and template.

If you want to verify that number yourself, read on.

How Azynote uses the Gemini API

Azynote makes Gemini API calls for four operations: OnePager generation, OneLiner generation, subject status, and semantic search indexing. Each is explained below with its cost.

What gets sent: the transcribed text from your session and any session notes. Audio is never sent anywhere. Azynote transcribes on-device with Whisper and discards the audio immediately. Only the resulting text goes to Gemini.

The four Gemini operations

1. OnePager generation

Your transcribed text, session notes, and the template instructions are sent to Gemini. It writes the structured OnePager.

2. OneLiner generation

Azynote reads the OnePager it just produced and asks Gemini for a one-sentence summary. This call uses the short OnePager as input, not the full transcript again, so it costs a small fraction of the first call.

3. Subject status

When you click "Generate status" on a subject, Azynote reads your most recent OnePagers for that subject (up to 5 by default, configurable in Settings → Sessions) and asks Gemini to write a short executive summary of where things stand. The output is 2 to 4 sentences. This happens only when you trigger it (manually or via Smart Sync batch), not per session.

Estimated cost on Gemini 2.5 Flash:

• Input: 5 OnePagers at roughly 2,000 tokens each, plus a small prompt, totaling about 10,500 tokens = $0.003
• Output: roughly 100 tokens = negligible
• About $0.003 per subject status refresh (a third of a cent)

4. Embeddings (semantic search)

When you turn on semantic search, Azynote indexes your OnePagers so that Chat and Search can find relevant passages by meaning, not just keywords. Indexing runs once per OnePager, then again whenever you update one. Raw transcripts are never indexed. Only the finished OnePager text is sent to Gemini, so no audio and no unprocessed session content ever leaves your machine.

Pricing for the embedding model (gemini-embedding-001) is $0.15 per million input tokens on the paid tier. A free tier exists with rate limits, which covers most individual usage at $0.

Estimated cost:

• Per OnePager (roughly 2,000 tokens): about $0.0003 (three hundredths of a cent)
• Indexing 100 OnePagers once: about $0.03
• Indexing 1,000 OnePagers once: about $0.30

Default setup

• Model: Gemini 2.5 Flash (the default, and the one used in all estimates below).
• Default template: Meeting Summary, which produces seven structured sections (agenda, key points, decisions, action items, open questions, next steps, and a brief conclusion).

Estimated cost by meeting length

These are estimates based on typical transcript density (around 130 words per minute) and the Meeting Summary template.

Input tokens include the transcript, your notes, the system prompt, and the template instructions (roughly 1,500 to 2,500 tokens combined). Output tokens include the OnePager body and thinking tokens.

The 1-hour meeting math

Google's published rates for Gemini 2.5 Flash (as of May 2026):

• Input: $0.30 per million tokens
• Output: $2.50 per million tokens (thinking tokens included)

For a 1-hour meeting with roughly 12,000 input tokens and 2,000 output tokens:

OnePager call:  12,000 × $0.30/M  =  $0.0036
                 2,000 × $2.50/M  =  $0.0050
                                     -------
                                     $0.0086

OneLiner call:  ~1,000 × $0.30/M  =  $0.0003
                  ~100 × $2.50/M  =  $0.0003
                                     -------
                                     ~$0.0006

Total: ~$0.009 ≈ $0.01

One cent. For a full hour of meeting capture and a polished written summary.

At 100 one-hour meetings a month, total API spend is roughly $1 to $2.

Typical monthly cost across all operations

Example: a user with roughly 1 hour of meetings per day, 20 working days per month, and 10 active subjects.

Chat usage is separate and depends on how much you type. A quick question is a fraction of a cent; a long back-and-forth with many context windows can add up. Watch your Google Cloud billing dashboard if you use Chat heavily.

Prices are verified as of May 2026. Check Google's official pricing page before making any budget decision.

The free tier

Google's AI Studio offers a free tier for Gemini 2.5 Flash with rate limits. Most individual users (a few meetings a day) stay within it comfortably. You get a free API key from aistudio.google.com, paste it into Azynote's Settings, and pay nothing until you hit the rate limit.

Rate limits reset daily. If you hit one mid-afternoon, Azynote will show an error on the next generation attempt. Switch to a paid API key if you consistently hit the limit.

Caveats

• Transcript density varies. A fast-paced technical discussion generates more words per minute than a slow exploratory conversation. Denser transcripts cost more.
• Templates differ in output length. The Meeting Details template produces a much longer OnePager (exhaustive summary, all people mentioned, references). Expect 2 to 3 times the output tokens of Meeting Summary for the same meeting.
• Thinking tokens add up with complex content. Gemini 2.5 Flash uses thinking tokens on harder tasks. Google bills them as output tokens.
• These are estimates, not guarantees. Your actual spend depends on your content. Check your Google Cloud billing dashboard if you want exact figures.

How to reduce cost

Stay on Flash. Gemini 2.5 Pro is roughly 4 to 6 times more expensive than Flash for typical Azynote workloads (which are well under 200k tokens). The same 1-hour meeting that costs about $0.01 on Flash costs about $0.04 to $0.05 on Pro. Flash handles meeting summaries well. Switch to Pro only if you have a specific reason (longer context needs, more complex reasoning).

Use the free tier first. For most individuals, AI Studio's free rate limit covers daily use. Only move to a paid key when you actually need to.

Pick a leaner template. Meeting Summary uses fewer sections than Meeting Details. For routine syncs, it is usually sufficient. Reserve the detailed template for important reviews or client calls where you need the full record.

Batch your generations. Generating five OnePagers in a row costs the same per-meeting as generating them one at a time. No penalty for batching, but no discount either. Just worth knowing there is no reason to delay.

Questions

Email support@azynote.com if you hit unexpected API charges or see errors you cannot explain. Include your Azynote version (shown in Settings) and the approximate time of the failed generation.