PLI Messages
Understand PLI (Picture Loss Indication) messages — what triggers full keyframe requests, why they impact video quality, and how to reduce them.
PLI stands for Picture Loss Indication. It is a message sent from a video receiver to the sender when the receiver has lost enough data that it can no longer reconstruct the current video picture. The receiver is essentially saying: "I have lost track of what is on screen — please send me a fresh keyframe so I can start over."
Think of it as tearing out a corrupted page from a flipbook and asking the artist to redraw the full scene from scratch instead of just the changes.
How It Works
Video codecs compress data by sending a full frame (a keyframe or I-frame) and then sending only the differences between subsequent frames (P-frames and B-frames). The receiver reconstructs the video by applying each difference to the previous frame. If a reference frame or a critical delta is lost and cannot be recovered via retransmission (NACK), the receiver's reconstruction breaks down entirely.
When this happens, the receiver sends a PLI message via RTCP feedback. The sender responds by encoding and transmitting a new keyframe — a complete frame that does not depend on any previous data. This allows the receiver to resynchronize and resume normal decoding.
Keyframe cost
Keyframes are significantly larger than delta frames — often 5 to 10 times the size. Each PLI-triggered keyframe causes a sudden bandwidth spike that can itself trigger further packet loss if the network is already constrained.
Why It Matters
PLI messages indicate a more severe problem than NACKs. While a NACK means a single packet was lost, a PLI means the receiver has given up on reconstructing the current picture entirely. Between the PLI request and the arrival of the new keyframe, the video is frozen or showing artifacts. Users see this as a visible glitch or pause.
Frequent PLI messages create a destructive cycle: each keyframe consumes extra bandwidth, which can cause more packet loss, which triggers more PLIs. In load testing, a rising PLI count is a strong signal that the system is approaching or exceeding its capacity.
Thresholds
| Level | Value (per second) | Interpretation |
|---|---|---|
| Good | 0 - 5 | Occasional recovery, minimal visible impact |
| Warning | 5 - 20 | Frequent picture loss, users will notice freezes |
| Critical | Above 20 | Continuous keyframe requests, video is severely degraded |
Common Causes of High PLI Counts
| Cause | Explanation |
|---|---|
| Sustained packet loss | Loss exceeding what NACK retransmission can recover |
| Burst loss | Multiple consecutive packets lost, destroying entire frames |
| Network jitter | Packets arriving too late for decoder to use |
| Encoder overload | Sender cannot produce keyframes fast enough under load |
| Insufficient bandwidth | Keyframe spikes consuming remaining headroom |
How to Fix It
- Address underlying packet loss — PLIs are a symptom of unrecoverable loss. Fix the root cause by following the guidance in the NACK and Packet Loss metric pages.
- Increase keyframe interval thoughtfully — A shorter keyframe interval means the receiver never has to wait long for a natural keyframe, reducing the impact of each PLI. However, shorter intervals also increase bandwidth usage.
- Reduce video resolution or frame rate — Lowering the encoding parameters reduces bandwidth consumption and the probability of loss.
- Ensure adequate bandwidth headroom — Size the link for peak keyframe bursts, not just average bitrate.
- Enable FEC (Forward Error Correction) — If your system supports it, FEC adds redundancy to the stream so the receiver can reconstruct lost packets without retransmission.
Related Metrics
- NACK Messages — The lighter recovery mechanism that PLI escalates from
- FIR Messages — An even stronger keyframe request for complete decoder state loss
- SLI Messages — Partial frame loss, less severe than PLI
- Freeze Count — The visible result of unrecovered picture loss
RFC Reference
PLI is defined in RFC 4585 (Extended RTP Profile for RTCP-Based Feedback), Section 6.3.1. It uses the RTCP Payload-Specific Feedback Message type (PT=206) with FMT=1. PLI carries no additional parameters — it is a simple signal that the receiver needs a new reference frame.
NACK Messages
Understand NACK (Negative Acknowledgement) messages in video RTP streams — what they measure, why they matter, and how to reduce retransmission requests.
FIR Messages
Understand FIR (Full Intra Request) messages — the strongest keyframe request mechanism, when it triggers, and what it means for video quality.