Does it make sense to already store the proposed libraries somewhere already?
Don’t you also say:
“El que no llora no mama”?
This is used in excess 
Let’s make the initial list, then let’s create the repos and start storing what is needed.
Regards
We should host files only if there isn’t a good public source. By “good” I mean it
- is feasible to script pulling files from
- has a means to provide feedback and get things fixed (e.g., Github repos)
Hi everyone,
I cheated a little and used AI to help me with the research. The results and the processing/setup approach are below.
I focused only on woodwinds (no brass, no saxophones). Even with that constraint, the research quickly points to a few dominant “source families” that come up most often and cover the majority of the common orchestral woodwind instruments.
Criteria I used (sfizz-only, factory-friendly)
Hard / must-have
- Clear license for samples and SFZ scripts (preferably permissive).
- Provenance: source of the recordings is documented.
- sfizz compatibility: SFZ features should be within what sfizz supports (avoid fragile ARIA-only extensions).
Soft / scoring
4) Coverage (sustain + short articulation / performance patches).
5) Reasonable defaults (ranges, levels, structure).
6) Practical footprint (size / maintenance / long-term clarity).
Recommendation table (per woodwind)
| Woodwind instrument | Availability in VSCO2 CE / SSO / VPO / FreePats | Recommended default source | Why it wins (for factory use) | Key risks / caveats | Second choice |
|---|---|---|---|---|---|
| Piccolo | VSCO2  / SSO / VPO / FreePats  |
VSCO2 CE | CC0/public domain baseline, low redistribution risk | May have fewer articulations than SSO | SSO (richer patching), but license risk |
| Flute | VSCO2 / SSO / VPO / FreePats (not orchestral) |
VSCO2 CE | CC0/public domain baseline, low redistribution risk | Possibly less “performance/keyswitch depth” than SSO | SSO (more articulation/KS), license risk |
| Alto Flute | VSCO2 / SSO / VPO (likely) / FreePats |
SSO | Explicitly included in SSO woodwinds | Sampling Plus + provenance concerns | VPO (bundle approach), but license mosaic |
| Recorder | VSCO2 / SSO / VPO / FreePats |
FreePats | Clear CC0 + per-instrument bank | Usually simpler articulation set | — |
| Oboe | VSCO2 / SSO / VPO / FreePats |
VSCO2 CE | CC0/public domain baseline, low redistribution risk | SSO may be richer in articulations | SSO (if acceptable), license caveat |
| English Horn (Cor Anglais) | VSCO2 / SSO / VPO / FreePats |
SSO | Explicitly included (Solo Cor Anglais) | Sampling Plus + provenance concerns | VPO, but license mosaic |
| Clarinet | VSCO2 / SSO / VPO / FreePats |
VSCO2 CE | CC0/public domain baseline, low redistribution risk | SSO may be richer; FreePats is simpler | FreePats (also CC0, small/clean) |
| Bass Clarinet | VSCO2 / SSO / VPO / FreePats |
SSO | Explicitly included (Solo Bass Clarinet) | Sampling Plus + provenance concerns | VPO, but license mosaic |
| Bassoon | VSCO2 / SSO / VPO / FreePats |
VSCO2 CE | CC0/public domain baseline, low redistribution risk | SSO may have more articulation | SSO (if acceptable), license caveat |
| Contrabassoon | VSCO2 / SSO / VPO / FreePats |
SSO | Explicitly included (Solo Contrabassoon) | Sampling Plus + provenance concerns | VPO, but license mosaic |
Practical takeaway for Zynthian factory curation
- Where available, VSCO2 CE is the safest default because CC0/public domain minimizes redistribution risk.
- For instruments that VSCO2 doesn’t cover (e.g., Alto Flute / English Horn / Bass Clarinet / Contrabassoon), SSO is the most complete SFZ set, but should be clearly labeled due to Sampling Plus + upstream provenance concerns.
- VPO is best treated as an optional add-on, not the legal baseline, because it aggregates multiple sources and therefore inherits a more complicated license/provenance situation.
- FreePats is excellent for small, clean CC0 single-instrument banks (notably recorder and clarinet).
Have a nice days.
2 Likes
When samplesets are in a repo and there’s a license file for the repo, we should assume (as everyone would) that the license applies to all files, not just audio files. This is industry standard practice. The only cases where we need to worry about sfz vs sample licenses is when they’re hosted separately. In that case there’ll be two source links, so we’ll have a clue. Whenever anyone handles one of these, if the licensing isn’t clear, mention it in the notes.
In a lot of cases we can probably get licenses for separate sfz files cleared up.
Regarding orquestral sounds, i don’t think we could have license issues with any of these 3:
- VSCO2 CE => VSCO 2: Community Edition (CE) - Versilian Studios
- SSO (sonatina) => GitHub - peastman/sso: Sonatina Symphonic Orchestra
- VPO3 => Virtual Playing Orchestra – Virtual Playing
As far as we don’t close or "sell* the resulting library, what we are not going to do, and we give credit to the original authors, nobody is going to feel upset because we build a library taking parts from these. Zynthian is a fully open software & hardware project and we are not doing something different to what others have already done (VPO3, for instance).
So, in this case, we should stick to quality and performance criteria for choosing. To ease packaging and maintenance, it would be desirable to use a single source for the factory library, but we could use a mix of them if it results advantageous, and it seems you think so.
VPO3 is currently included in zynthian because it’s already a “mix of the best” of VSCO2, SSO and others. Of course, the “best of” criteria is a question of taste, but i considered that the authors of VPO3 would have better criteria than me and they have dedicated more time and effort to do the selection that i could. Of course, if you think we could do it better, please, go forward and let’s build the ZVO or what ever we call it 
Best Regards,
My personal opinion is that the nature of these libraries partially being remixes of each other and varying different approaches of scripts towards similar / shared sample sets should encourage us to make the best version for an integrated keyboard extender we can do.
This is the way.
I have made a couple of changes to vangelis webconf. You can now upload 7z and rar packaged presets to fluidsynth and sfz. You can do free search as well as tag search in music-artifacts. (Previously it was only tag search which tended to give poor results.) I also improved the certificate handling a little which increases the chances of a query working. These changes should improve the ability to search and install from musical artifacts.
2 Likes
The author of VSCO2 is active on the sfz discord, so if we have any questions or requests LMK and I’ll be happy to ask. As you can imagine, he’s very community-minded and also very knowledgeable about sampling in general and orchestral instruments in particular. I think he’d be in favor of any use of the samplesets he’s released as open being used on Zynthian.
I don’t generally use orchestral samplesets, but in the VSCOx samplesets I’ve looked into, I’ve found nothing to complain about. They’re all really very good. But I don’t recall all the license details, which I’ll review in a few days.
I upgraded Oram to Vangelis a few days ago, but sadly it doesn’t work at all for me. I don’t have any Zyn hardware, just mouse and monitor. When trying to create or edit a chain, when I click on the chain name at the top of the screen, the UI locks up. ( Clicking on chain name locks up UI · Issue #1585 · zynthian/zynthian-issue-tracking · GitHub )
What’s the best way to update to latest test/Vangelis – use the webconf and do the same thing I did before, or use apt update ? I’m hoping it’s a temporary failure. I really want to try the new CC value mapping feature!
@hannesmenzel What’s the purpose of the numbers in each line in the tables in the wiki?
In a couple days I plan to jump in on that. My first step would be to fill it with what we currently have, and noting any license issues. I’ll also identify them somehow as currently loaded samplesets.
I suggest that whenever someone adds lines that are preliminary suggestions that haven’t been vetted musically, we mark them as “preliminary” and encourage anyone with an interest to check them out.
Once we have a good list, we can ask the broader community to review and make suggestions. Something that would really help is an audio demo file for each one. Yeah, that’s a lot of work, but I think it’d be worthwhile. (We could also make examples available via the UI, if we do it right.)
I think I can manage to get audio examples added to repos for every sampleset at sfzinstruments · GitHub .
1 Like
Do not manually use apt. It can lead to issues. Use the software update option in webconf of zynthian admin menu.
OK, no apt, thanks!
PS: Can someone please create an issue for this work? Of course there will be a number of sub-issues to do it all.
I’d also like input as to whether to use the existing zynthian-data repo for hosting full samplesets (including audio files) or whether we’d want a separate repo for each one, etc. I think that zynthian-data should be good for cases where we supply only YML and/or SFZ files. Please LMK if you think otherwise.
Yes, I think we intend to use the samplesets as open. (crediting author and source) In terms of modifications, the intention of the creator is reflected in the CC-0 licence. I have nothing to complain about the sets either, but in terms of usage in Zynthian I have complaints about the folder structure, the sfz file per articulation and the fact, that I’d like some more performance parameters.
1 Like
7 posts were merged into an existing topic: Zynbleton Vangelis - READ ME
I updated the wiki table at Refactoring Zynthians factory soundfont library - ZynthianWiki , adding grand pianos, Rhodes, steel drums, and marimba.
I also updated to use the following columns:
- instrument
- name and link to source
- license
- editor
- primary/secondary (primary are distributed in the image; secondary would be easily added by the user.) Any sources with questionable licenses are secondary. We can also use this for large or somewhat redundant but possibly popular options.
- notes
I pointed out that it’s silly for us to have sections for synths and drawbar organs, since those bases are already covered so much better by actual synths and SetBFree. But I didn’t delete them yet. Opinions?
BTW, I don’t recommend we necessarily use the name in the “instrument” column in the UI. I’d rather name them explicitly, so anyone familiar with the source knows what it is. But also, where appropriate, making it clear what it actually is. For example: “jRhodes3c - looped stereo Rhodes” rather than just “stereo Rhodes” in the UI.
Are we worried about compatibility? If someone upgrades, won’t they lose all their saved chains, if we make radical changes to how these are organized?
I’m hoping we can find a way to manage this so that upgrades don’t ruin saved chains. If we’ve moved where samplesets go, we can keep links from the old locations to the new ones (and possibly, a “legacy” switch in the UI to show the old ones.)
If we’re removing samplesets with questionable licenses (and I think we should) we can provide a way for the user to load those. Ideally, when loading chains, if a removed sampleset appears, we can provide instructions on how to resolve the issue. Perhaps we could also have a “chain checker” that looks for issues in all snapshots, that can be run as part of the upgrade.
Can someone please point me to the source in our repos where we load samplesets from the original sources?
I wouldn’t delete them now from the table, because either we find some nice patches or we don’t, and in this case they will not find their way into the final folder structure. I do agree for the synth, but there was a discussion in another thread recently about the fact you can only have one instance of SetBFree, and the given solution there was to use soundfonts if you need more than that.
I think sources with questionable licenses are out.
I totally comply with that. I would switch it ("Rhodes - jRhodes3c) for alphabetical ordering purpose (P.S.: I’m not sure but I think stereo Rhodes is a different sample set from yours).
Great you continued that! We might need to fix the table formatting.