Freezetime vs SC2 Replay AI Coach

Freezetime is a free, self-hosted analysis platform for Counter-Strike 2 demos. Its public archive is a read-only site with more than 2,600 professional matches, searchable and grouped by series, and needs no installation or account. For a personal demo, the Analyze page parses one file in the browser using WebAssembly, so the file stays on the user's machine. Processing produces 16 Hz player positions along with kills, grenades, and economy data. Synchronized 2D replays show player health, economy, kill feed, grenade trajectories, smoke, and flash effects. Analysis tools include opponent reports, predictions, ML Lab, leaderboards, match comparison, heatmaps, ghost rounds, Moments, and Pattern Finder. My DB can process private demos without the server keeping a copy. Self-hosting unlocks features marked self-host only, but users must supply their own demos because the repository contains no match data or database dumps. The project says smart features use deterministic methods and locally trained models rather than external AI services. Freezetime is MIT licensed, with no warranty.

SC2 Replay AI Coach is a free browser-based site for parsing, viewing, and analyzing StarCraft II and Brood War replay files. It accepts .SC2Replay files for StarCraft II and .rep files for Brood War. Replay pages present economy timelines, army composition, battle records, build-order timelines, strategy classifications, and coaching notes. The automated post-game coach reads replay events, derives economy and combat metrics, compares openings with known build patterns, and adds notes tied to replay timestamps. Deterministic rules identify tactics including drops, cannon rushes, proxies, fast expansions, and one-base all-ins. Uploaded replays remain available through shareable URLs, with listed team coverage from 1v1 through 4v4. The tool requires no account or download and states no replay-count limit. Uploading is available in desktop browsers where local replay files are stored. Co-op replay visualization is supported, but ladder-coaching assumptions do not fully fit commander-specific units and mission objectives. Frame-perfect micro grading is limited because replay data contains events and sampled positions rather than a full coaching video.