Splice FAQ: privacy, cost, AI models, and clip storage

Answers about Splice privacy, OpenRouter costs, supported capture tools, AI models, Explore expiry, and where your clips are stored.

Does Splice upload my clips?#

Sometimes — to OpenRouter, for analysis. For most built-in models (Gemini 3 Flash, Gemini 3.1 Flash Lite, Qwen 3.6 Plus, Nemotron 3 Nano Omni) Splice uploads the extracted highlight directly so the model can sample it. For image-only models (the Gemma family), Splice extracts JPEG frames at your configured FPS and sends those instead. Either way, your full source recording never leaves your PC. The only time a clip leaves your machine in a publicly-accessible way is when you explicitly hit Publish to post it to Explore.

Does Splice cost money?#

Splice itself is free. AI analysis uses your own OpenRouter account, so you only pay for what you actually run. A typical 30-minute gaming session costs a few cents at the default settings with the default model. You can see the exact cost for each job in the queue after it finishes.

Do I need an NVIDIA card?#

No. Splice works with any capture tool that saves video files to a folder — NVIDIA Instant Replay, AMD ReLive, OBS, Xbox Game Bar, Medal, or anything else. If you’re already saving clips somewhere, point Splice at that folder and you’re set.

Will Splice slow down my game while I’m playing?#

No. Splice waits until your game’s process exits before kicking off analysis — alt-tabbing isn’t enough; it watches for the game’s .exe to close. The folder-watching and queue management that happen in the background use almost no CPU or RAM.

Can I delete the highlights Splice made?#

Yes. You can delete any highlight from the Highlights page, or from the recap dialog on the Dashboard. Deleting permanently unlinks the file from disk — it does not go to your Recycle Bin. Your original recording (the full source clip from your recorder) is never affected. Deleting a block from inside the timeline editor only removes it from that project; the underlying highlight file stays put.

What happens to my clips after I trim them?#

It depends which trim you mean:

• The Edit Kill Detection dialog on the Dashboard re-extracts the clip with FFmpeg’s stream-copy path — nothing is re-encoded, the new file is byte-identical quality. Saving here always replaces the existing highlight; there’s no keep-both option.
• The timeline Editor on the /editor page re-encodes once when you click Render (it has to, because it mixes audio and stitches multiple clips together). The render pulls from the original source recording, so re-rendering doesn’t introduce generational quality loss.

Either way, your original recording is never modified.

What if Splice misses a highlight I know I got?#

A few things can cause misses: your sample rate might be too low, your detection prompt might be too strict, or the model you’re using might not be confident enough for that scenario. Open the Models page, expand the game’s card, and try bumping its FPS slider, loosening the detection prompt, or picking a more capable model. You can also re-run analysis on the same clip from the Reanalyze menu on the Dashboard.

What happens to a clip after it expires from Explore?#

After 14 days, the clip and thumbnail are automatically deleted from Splice’s servers. The share link shows “not available.” Your local copy is completely untouched — it stays in your highlights folder as long as you want it.

Can I run Splice on macOS or Linux?#

Not yet. The current release targets Windows 10 and 11. macOS support is on the roadmap — the core tech works cross-platform, it’s just a matter of packaging and testing. There’s no Linux timeline yet.

Is the source code open?#

Parts of it. The desktop app shell, the clip pipeline, and this marketing site are public. The moderation pipeline and a few internal services are not.

What if I want to use a different AI than the ones listed?#

Splice ships with six built-in vision models (Gemma 4 31B, Gemma 4 26B A4B, Qwen 3.6 Plus, Gemini 3 Flash, Gemini 3.1 Flash Lite, Nemotron 3 Nano Omni). You can add any other vision-capable OpenRouter model from the Models page using the Add Custom Model form — type its model ID and pricing and it shows up in the picker. New OpenRouter models don’t auto-populate from their catalogue; you add them by hand. If you want to use AI from a different provider entirely, that’s not supported yet — OpenRouter is the only AI integration for now.