Recent comparisons from one of our forum users.

The following was recently posted by Rich on the VSee forum.

Although I’ve not been posting much lately, I am a constant and enthusiastic tester of video conferencing applications. I’ve just completed some rather lengthy comparisons between some of the more common competitors.

method: two network conditions (office LAN with 20 Mbs in/out to our colo and access to Internet2 from there & commodity DSL max outbound about 300 Kbs) both permutations tested. Identical machines (Dell, Latitude D630 w/Centrino Duo & 2 MB RAM) Logitech, QuickCam Pro Notebook webcams; & either Plantronics .400 DSP headset or Phoenix Audio Duet PCS HW-AEC USB speakerphones

ooVoo 2.1.00.51 vs. VSee 9.0.0.416

1.) Tightest Lip-Sync (facial word-formation/utterance synchronized to heard speech) perceived by trained recipients: VSee
2.) Least amount of Smooth-Motion Distortion (Jerkiness) perceived by trained recipents: VSee
3.) Least amount of Block/Tiling Distortion & Mosquito Noise perceived by trained recipients: VSee
4.) Least amount of Object Persistance (lagging images from previous frames seen as faded or outline images on current frames) perceived by trained recipients : VSee
5.) Least amount of Scene Cut Response time (Time required to build up new image) perceived by trained recipients: VSee
6.) Highest: Edge-Sharpness & Spatial Detail perceived by trained recipients: VSee (ooVoo was a smear-fest)
7.) Least amount of Color Errors perceived by trained recipients: NA (for 1/2 the test participant was in too-dark room)
8.) Highest Voice/Speech PESQ/MOS perceived by trained recipients: ooVoo (Siren14 > SpeeX-wb)
9.) Least amount of required in/out Bit Rate for each participant: VSee
10.) Least amount of required End-Point Systems Resources: tie
11.) Best traversal of: Proxies, Firewalls, Symmetric & Full-Cone NATs, Packet-Shapers, etc.: tie

SightSpeed build 6564 vs. VSee 9.0.0.416

1.) Tightest Lip-Sync (facial word-formation/utterance synchronized to heard speech) perceived by trained recipients: tie
2.) Least amount of Smooth-Motion Distortion (Jerkiness) perceived by trained recipents: VSee
3.) Least amount of Block/Tiling Distortion & Mosquito Noise perceived by trained recipients: VSee
4.) Least amount of Object Persistance (lagging images from previous frames seen as faded or outline images on current frames) perceived by trained recipients : VSee
5.) Least amount of Scene Cut Response time (Time required to build up new image) perceived by trained recipients: VSee
6.) Highest: Edge-Sharpness & Spatial Detail perceived by trained recipients: VSee (SightSpeed is very “soft” a little smeared)
7.) Least amount of Color Errors perceived by trained recipients: tie
8.) Highest Voice/Speech PESQ/MOS perceived by trained recipients: tie – SightSpeed uses narrowband, but the comprehension of consonants and close sounding vowels was better in SightSpeed than VSee’s SpeeX-wb; or it’s narrow-band TrueSpeech
9.) Least amount of required in/out Bit Rate for each participant: too close to call
10.) Least amount of required End-Point Systems Resources: too close to call
11.) Best traversal of: Proxies, Firewalls, Symmetric & Full-Cone NATs, Packet-Shapers, etc.: tie

VZOchat 6.3 vs. VSee

VZOchat wasn’t even worth testing – it seemingly has no control over its packets – they go (or don’t go) willy nilly with no attempt at synchronizing audio and video that I could tell. Unusable (really we only tried this as it advertised multi-participants)

Vidivic 3.1.0.6 vs. VSee

Again, don’t bother

We’ve not tallied our subjective scores (1-5; five being the best) but these are the summaries. I did a similar test a year or so ago; same results really.

Comments ( 2 )
  • admin
    Rich says:

    One note on these comparisons – the area where VSee falls a bit short is in voice quality. Add SpeeX-WB has helped a lot; but I suspect it tops out at 7KHz. Audio CODECs like Skype’s SILK and Xiph’s CELT are great; but their bandwidth typically exceeds most microphone/speaker combos in use by most users. If, perhaps, SpeeX could be widened to 70Hz-14KHz (like Siren14 – a.k.a. G.722.1 annex C) the computational requirements don’t scale as much as the quality of the voice – all while keeping the bandwidth in the range of most consumer audio devices. I seem to recall SightSpeed (ne CU30) using some narrow-band CELT-based CODEC, but articulation seemed more precise.

    Having said this, I’d be happier with narrow-band in VSee if extending voice bandwidth in any way jeopardized lip-sync/smooth-motion quality. For those “in the know” lip-sync/smooth-motion is the quality that best delineates VSee from any other competitor (IMHO).

  • admin
    Rich says:

    edit to the original post – the Dell 630’s have 2 GB RAM, not 2 MB 🙂

Leave A Comment

Your email address will not be published. Required fields are marked *

%d bloggers like this: