Cinelerra: Cinelerra Video Editing tool quick start summary

This post was written by eli on January 1, 2009
Posted Under: Software

Ehm.. This is really outdated

Cinelerra has made a long way since this was written. So really, take this with a grain of salt.

Making the Dyne::Bolic live CD feel at home

  • Nest (under /dyne). 128 MB should be enough
  • Dock
  • Boot from hard disk
  • Make a swap partition (if it exists, it will be used automagically)

Starting a project

  • Start up cinelerra from fresh
  • Set the desired output format wisely (Settings->Format…). If you have a trancoder utility (e.g. mencoder), the most used input format is preferred. You are going to transcode the output to the target format afterwards anyhow.
  • Set the frame rate. Changing it later may disrupt A/V sync.
  • No less important: Check the resources’ info (right-click the video file’s entry in the resource in the resource window) and make sure that the frame rate matches the real frame rate exactly. A bug in Cinelerra may change this value slightly, causing A/V sync problems during rendering.
  • Set the video track(s) size to match the output’s: Right-click the video track and pick “Match output size” (or “Resize track…”). This is important in particular when editing PAL videos.
  • Load the video/audio files (File->Load files). “Create new resources only” is preferred. Take them all at once, since this can take time.
  • … but not too many: If Cinelerra crashes, take them a few at a time.
  • Set the preferences NOT to play every frame (Settings->Preferences->Playback->Video Out->Play every frame unchecked).
  • Also consider turning off Settings->Preferences->Playback->Audio Out->View follows playback if the video playback freezes often.
  • Save your project file (File->Save As…)

Input formats

  • Rule #1: Don’t lose your sanity
  • Cinelerra isn’t very friendly in this respect, but formally supports most things you’d like to import. Nevertheless, it’s very sensitive to quirks in the stream.
  • MPEG-2 (the format used in DVDs) is OK, but only for short clips (four minutes?) or A/V sync is slowly lost.
  • AVI/DIVX seems to be a good format for import
  • When using mencoder to create an input file, always use the -vf harddup option, so duplicate frames are explicitly coded in the stream. Failing to do so will make some keyframed effects go crazy. This is important in particular when changing the frame rate.
  • MJPEG works well. mencoder can be used for recoding (keep in mind that MJPEG involves some flicker noise): mencoder input_file.avi -oac mp3lame -ovc lavc -vf harddup -lameopts abr:q=5 -lavcopts vcodec=mjpeg -ffourcc MJPG -o output_file.avi
  • Videos marked as “divx” (as opposed to “DIVX”) are not recognized.
  • MPEG-2 is accepted, but a TOC file is generated in the .bcast directory, and that takes time.
  • Better alternative (when it works): mpeg3toc -v /path/to/video.mpg video.toc Always make the path absolute, since this path is written in the TOC file. Then load the .toc file (not the .mpg) to Cinelerra.
  • To get an MPEG-2 file from DVD track 1: mencoder dvd://1 -dvd-device /dev/scd0 -vf harddup -oac copy -ovc copy -of mpeg -mpegopts format=dvd -o video.mpg
    (Note that -dvd-device should point at your DVD player)
  • For really long MPEG-2′s, concatenating VOBs (with “cat”) is safer.
  • Other MPEG flavors should be OK as well (haven’t tried)
  • MJPEG from a Canon S3 camera works 100%

Two-screen editing

  • Browse through the media in the Viewer. Find the juicy snippets and send them into the timeline or make clips
  • See what you got in the Compositor
  • Labels are very useful for navigating in the timeline
  • Except for time jumps, what you see in the Compositor is what you’ll get.

Editing by draggin handles

  • In “Cut and Paste mode” (“I” toolbar button pressed), dragging the beginnings or ends of edits.
  • Only armed tracks are affected
  • With left button, edits are shortened or extended the overall time changes
  • With middle button, only the cut time changes, one edit “stealing” time from the other.
  • With right button, overall time is not changed, but media is shifted in the opposite direction of the drag direction. In effect, you mark the beginning or end of the edit, and it’s shifted to match that.
  • These are default mouse button settings, and can be changed.


  • The red button
  • Shift-click for arm one track only or all tracks

Add an image

  • Due to a bug in Version CV 2.1 (and probably others), this doesn’t work at all when the frame rate divided by 3 is an integer (or so it seems). Avoid the frame rate 30 (as opposed to 29.97)
  • Transparency is respected. Use it (in PNG images).
  • … but if there is nothing behind the transparent image, the background becomes white (or black. Or anything)
  • Load the image file as a resource
  • Create a video track
  • Select entry poiny in timeline
  • Zoom in timeline, so that one frame period is viewable
  • Disarm all tracks, arm the new track
  • Drag the image into timeline or right-click-paste on resource
  • Zoom out timeline and extend the duration of the image using edit dragging
  • See my bug report: CV version bug #459


  • Translate
  • Chroma Key / Difference Key
  • ReframeRT (slow motion / crazy motion)
  • Reroute (future versions)
  • Effects work on a single track
  • … unless they are explicitly shared


  • Close the compositor video, if you don’t want to watch the rendering process. (May slow down rendering significantly)
  • DO NOT RENDER TO AVI. Use Quicktime for Linux instead.
  • Heavy A/V sync problems have been experienced when rendering to AVI
  • To get a DIVX: Use MP3 audio compression (256 kbits?) and MPEG-4 Video with default parameters.
  • Then convert to AVI: mencoder -oac mp3lame -ovc copy -ffourcc DIVX -o clip.avi
  • Re-encoding to DIVX is advisable to be sure to get a clean file.
  • If A/V sync problems occur, render audio and video separately, and then multiplex. mencoder’s -audiofile parameter may come handy.
  • Note that it’s “DIVX”, not “divx”

Rendering for DVD

  • Video and Audio are rendered separately
  • MPEG-2 streams can be generated directly from Cinelerra, but the manual doesn’t tell you to do that.
  • … maybe because it’s easy to get off-standard with many GUI parameters to set.
  • For video, the output file format is YUV4MPEG, and check “To Pipe”.
  • Create an executable script (with #!/bin/bash) file saying: mpeg2enc -v 0 -K tmpgenc -r 16 -4 1 -2 1 -D 10 -E 10 -g 15 -G 15 -q 6 -b 8600 -f 8 -o /path/to/output-video-file.m2v
  • Note that unlike what’s implied in the manual, Cinelerra doesn’t pass the output file name as first argument to the script. The file name is therefore hardcoded
  • It does send some garbage instead. A simple trick is to end the pipe’s command with a ‘#’ which comments out the garbage
  • Then choose AC3 as output format, and render audio. The standard allows up to 1536 kBits/sec, so you can’t go wrong with this.
  • Note that you can use mencoder and YUV4MPEG as well for a “safer” encoding.
  • Check the console for error messages and progress (when using mencoder)
  • When using mencoder, create a raw PCM audio file (Microsoft WAV is fine) and then apply the file with the -audiofile parameter.
  • In the script file you’ll have something like: mencoder -audiofile "full/path/to/" -oac lavc -ovc lavc -of mpeg -mpegopts format=dvd -vf scale=720:480 -srate 48000 -af lavcresample=48000 -lavcopts vcodec=mpeg2video:vstrict=1:vrc_buf_size=1835:vrc_maxrate=9800:vbitrate=5000:keyint=18:acodec=ac3:abitrate=192:aspect=4/3 -ofps 30000/1001 -o "movie.mpg" - (NTSC)
  • Note the dash, telling mencoder to read from standard input
  • If mencoder complains that it “cannot seek backward in linear streams” and “seek failed” (several times), adding the -cache 8192 parameter is likely to help.

General remarks

  • Cinelerra crashes a lot. Save your project file frequently.
  • Actually, it’s recommended to save, quit and restart every 15 minutes or so (until you learn how to avoid messy situations).
  • Like any typical video application, Cinelerra is pretty buggy.
  • Not a single byte of media is written until you render. All edits are merely pointers to existing media
  • … therefore, never move, change or delete the media files, once loaded to Cinelerra.
  • Cinelerra creates backup.xml index files (and other things in) ~/.bcast
  • The *.rc files in ~/.bcast are the defaults for plugins. When these are getting annoying, remove this directory between Cinelerra sessions
  • DO NOT render output AVI files
  • Always have Settings->Align cursor on frames checked when editing video
  • Do not use Copy/Paste/Cut/Delete as you usually would in a text document. They will appear to work, but trouble will soon come. Use two-screen editing functions, drag handles or completely remove tracks instead.
  • … except for copy-pasting a piece into a fresh, second Cinelerra instance for integrating several scenes to a movie (or just cleaning up the environment).
  • In/out points can be applied on the main timeline. This can be used to create a clip or render a part of the full video, but be sure to remove these when not used. In particular, inserting material from the Viewer can fail (due to a bug) if In/Out points are marked on the timeline (as opposed to In/Out points on the Viewer, which you can’t without).
  • Having too many “resources” (assets, things in your media list) can make Cinelerra unable to load new ones. Cinelerra keeps up to four file handles open for each, so when attempting to load another one, Cinelerra hangs and an error message misleadingly implying that the file is missing may appear at console.
  • In the EDL (that is, the XML file containing edits), the timeline is given in integer frames for video edits, transitions and keyframes, based upon the expected output frame rate. This means that changing the output frame rate will cause some inprecision (in particular 30 to 29.97).
  • Same applies for output audio sample rate with respect to audio edits etc.
  • Output with alpha channels (transparency) is possible, for example as Quicktime for Linux with Uncompressed RGBA, or PNG output sequence. If the overall rendered image has some transparency to it, that will be reflected in the output’s alpha channel. There are only two things to note: One is that sometimes a black and completely transparent image must be put at the bottom of the timeline (master background) or elements, which should have been transparent or faded away show up. The second issue is that if non-normal overlay is used (such as multiplication overlay), the alpha channel will be set to opaque at the output.

Video formats (more…)

  • For intermediate video output: Use Quicktime For Linux / Uncompressed RGB

Add a Comment

required, use real name
required, will not be published
optional, your blog address