<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 TRANSITIONAL//EN"> <HTML> <HEAD> <META HTTP-EQUIV="Content-Type" CONTENT="text/html; CHARSET=UTF-8"> <META NAME="GENERATOR" CONTENT="GtkHTML/4.6.6"> </HEAD> <BODY> Le mardi 13 mai 2014 à 09:17 -0300, Thiago Santos a écrit : <BLOCKQUOTE TYPE=CITE> hlsdemux doesn't need to drain downstream demuxer/decoders before doing the switch, so it seems something else would be the issue. Is this happening with every stream or just with some particular one? The buffering inside the pipeline should be enough to allow the switch without causing pauses, even if there is some overhead. Does this happen both when switching to a better and to a lower resolution? </BLOCKQUOTE> Note, it depends a bit on the setup. If sink is offering the pool, it means the allocation query will reach the sink before returning (during renegotiation). This is in fact a drain. To be gapless, a decoder will have to produce 1 decoded frame before deadline for the next frame is reached. Ideally, we have to try and make the decoder src pad renegotiation happen has late as possible, to reduce that time. This is something worth tracing. Nicolas </BODY> </HTML>