Comparison
Frameskim vs Gemini: frames as image files, or a text answer
Gemini reads a public YouTube URL and answers in text. Frameskim returns the frames as image files, with their timestamps.
Short version. Use Gemini if you only need a text answer about the video. Use Frameskim if you need the real image at the exact second, or if your model is not Gemini.
Side by side
| Gemini API (YouTube URL) | Frameskim | |
|---|---|---|
| Input | A public YouTube URL in the prompt. The feature is in preview. | A public YouTube URL, one video per run, via Apify API, MCP or n8n. |
| Output | Generated text. Not images. | JSON with image URLs, transcript segments and moments. |
| Frames as image files | No. The model sees them, you do not get them. | Yes. One image per frame, at your interval. |
| Timestamps | References in MM:SS written inside the text answer. | The time of each frame, as a field of each item. |
| Transcript | Not a separate output: ask in the prompt, get generated text. | Caption segments aligned to each frame. Needs captions on the video. |
| Works with | Gemini models. | Any model that takes images: OpenAI, Claude, open models. |
| Cost per 10-minute video | Free today with a YouTube URL (preview, "no charge"). Per token for uploaded files: $0.006 to $0.075, see below. | $0.42 for 20 frames. Pay per event, no subscription. |
The OpenAI and Claude vision guides describe image input. With those models, the frames have to come from somewhere.
The cost math
With a YouTube URL, Gemini is free while in preview ("no charge"). Per token applies to uploaded video files or after the preview ends.
Gemini: about 100 tokens per second of video
| Model | Per 1M tokens | 10 min, 60,000 tokens |
|---|---|---|
| 2.5 Flash-Lite | $0.10 | $0.006 |
| 2.5 Flash | $0.30 | $0.018 |
| 2.5 Pro | $1.25 | $0.075 |
Standard tier prices listed 2026-10-09. Output tokens extra. Sources below.
Frameskim: $0.12 per video + $0.015 per frame
| Event | Count | 10 min, 20 frames |
|---|---|---|
| Video processed | 1 | $0.12 |
| Frames with image | 20 | $0.30 |
| Total | $0.42 |
Same example as the home page. You pay Frameskim for the image files.
Who should use which
Use Gemini if
- You only need a text answer about the video.
- Your stack already runs on Gemini.
- You want the lowest cost per video for understanding it.
Use Frameskim if
- You need the real image at the exact second: for a document, a ticket, a blog post.
- Your model is not Gemini: OpenAI, Claude, an open model, or n8n without video support.
- You want evidence you can check: each frame comes with its exact time, so you can check it yourself.
Pay per event on Apify. No subscription.