Everything you need to know about using PitchView AI's video analysis platform
Step-by-step help for every part of PitchView AI, based on how the portal works today.
From sign-up to your first timeline of events in five steps:
Sign up on the PitchView AI portal with your email address and a password, then sign in. Your profile shows how long you have been a member and the totals of your analyses and detected events.
Before your first analysis, add the teams you work with on the Teams page. When you start an analysis you can search for them and select them as Team A and Team B, so events are credited to the right side.
If you want the AI to follow your own instructions, create a custom prompt on the Prompts page.
On the Analyze page, pick the source type. For a YouTube link you can also choose which video quality to use. Then select the sport and, optionally, the teams and the prompt.
Once started, the analysis moves through clear stages β downloading the video, extracting its metadata, uploading it to cloud storage and AI analysis. A typical video is done in 2 to 5 minutes.
For full-length matches, use Parallel: it splits the video into segments and analyses several of them at the same time (three by default), so long videos finish much faster.
Each result is a timeline of events. Every event has a timestamp, an event type (for example Goal, Foul or Corner), a description, the team involved and its source β AI or manual.
Click any timestamp to jump to that moment in the video. You can search the events and filter them by type, review status and team.
You stay in control of the data: mark each event as Correct, Wrong, Uncertain, Partial or Duplicate (new events start as Pending) with an optional comment, edit or delete events, add events the AI missed, or run the whole analysis again with Reanalyze.
You can upload MP4, MOV, AVI, MKV and WebM files. Any other file type is rejected with a message listing the accepted formats.
You do not have to upload at all: paste a YouTube link or a direct link to a video file and PitchView AI fetches the video for you.
Paste a YouTube link on the Analyze page. PitchView AI reads the qualities the video is available in so you can choose one, and it can detect the sport from the video automatically.
The video is downloaded and kept in cloud storage for the analysis, so analysing the same video again is faster.
Only videos PitchView AI can access can be analysed β private or removed videos cannot be downloaded.
Each uploaded file can be up to 500 MB.
If your recording is larger, analyse it from a YouTube link or a direct video URL instead of uploading the file, and use Parallel for full matches.
Most upload problems have a quick fix:
PitchView AI sends your video to Google Gemini multimodal models together with a prompt written for the selected sport. The model watches the video and returns the events it finds, each with a timestamp, type, description and team.
The events are saved with the analysis, so you can review, correct and export them. Long videos can be split into segments that are analysed in parallel.
The prompt tells the AI what to look for. PitchView AI includes system prompt templates for each sport, and on the Prompts page you can create your own: give it a name, write the instructions and use it for your next analyses.
Custom prompts are useful when you care about specific moments β for example only set pieces, or the actions of a particular player.
Advanced users can tune the AI model from the Prompts page. Without custom settings, analyses use the default settings, and the settings used are shown in each analysis in History.
AI detection is strong but not perfect: it can miss an event, misread one or report the same moment twice. That is why every event can be reviewed.
You get the most accurate results by choosing the right sport, adding both team names, using a clear, good-quality video and keeping prompts focused.
The review statuses β Correct, Wrong, Uncertain, Partial and Duplicate β give you a clear record of how reliable each analysis was, and they are included when you export.
While a video is being analysed you see each stage update live β download, metadata, cloud upload and AI analysis β without refreshing the page.
For live matches broadcast with PitchView, Stream analyses the broadcast while it plays: you set the analysis window in seconds and the maximum number of minutes, and detected events appear under Live events as the match goes on.
Every analysis is saved in History, with the total number of analyses and events and your most-analysed sport.
Search by team, ID or content, filter by sport and status, open any analysis to review its events, run it again with Reanalyze, or delete analyses you no longer need.
Inside an analysis, events are listed in time order. Click a timestamp to jump straight to that moment in the video.
Search events by type, description or team, filter them by type, status and team, and add a manual event at any time β manual events are labelled so you can tell them apart from AI events.
Use Export on any analysis to download its events as an Excel (.xlsx) file.
Each row contains the time, the time in seconds, event type, description, team, source (AI or manual), review status, when and by whom the status was changed, and the event ID β ready to share with coaches and analysts or to load into your own tools.
API access is included in the plans that list it on the pricing page.
Authentication details, the endpoint reference, rate limits and code examples are provided when API access is enabled for your account. Contact us to get started.
We support MP4, AVI, MOV, MKV, and WEBM formats. Maximum file size is 5GB per video.
Analysis typically takes 2-5 minutes depending on video length and complexity. You can watch progress in real-time.
Yes, you can analyze any YouTube video as long as you have the link, including unlisted videos.
Our AI achieves 99% accuracy in event detection across all supported sports, continuously improving through machine learning.
Yes! You can create custom prompts and configure LLM settings to tailor the analysis to your specific needs.