Posted
As I get more comfortable with Synfire, and my libraries grow, I wonder about how to produce the best quality figures and feel I'm not really adept at optimizing figures for reuse. If, when importing a midi file, for example, a hit pop song, do I need to be faithful at the harmonization stage. For instance, if I want to preserve the actual original 'ideas' at play with each instrument, and the dynamic between them over time, should I go through and enter the original chord progression at the harmonization stage, or just let the harmonization do it's work, and sort of select chords by ear. Worried that the integirty of the 'ideas' will be compromised if the wrong chords are chosen. i.e the figures will not mimic whats happening. What approach do others take when importing for figure reuse, especially over time (with respect to what happens at transitions in the structure), and the dynamic between instruments. Any suggestions be great. I really want to capture the 'ideas' at play and imitate it with my music.
Mo., 30.11.2009 - 12:23 Permalink
well put. probably an easier question to answer would be 'what is the meaning of life?' remember Andre mentioning the only problem you'll have is deciding on one of the infinite possibilities.
guess i'm more concerned with capturing the change/movement in regards to transitions, harmony, etc and learning from and imitating. is it best to import a whole song as an arrangement, sort of keep it in arrangement form as a template, pick what you want, optimize and apply it to my own composition, then tinker with it make it fit, hopefully preserving the idea?
Mo., 30.11.2009 - 13:18 Permalink
Undoubtedly, the better an import/take is harmonized, the better your automatically recognized figures will be and the less manual editing is needed to make them correctly mimic what's happening.
You immediately see this when the resulting figure looks "minimal" and doesn't show sudden leaps where they don't belong.
Therefore, if you know the chords already (e.g. from a song book), use them! If the harmonizer doesn't have them in its suggestion list, you can copy/paste from a palette or a progression editor (where you can type them in).
The workflow is:
1) Harmonize your static import/take
2) Tweak recognition settings until you see a good looking figure
3) Edit the figure to behave nicely in different contexts
The Phrase Pool Editor is designed for this task. I just reviewed it and fixed a couple rather serious problems, especially concerning (3) (will be included in next build).
It's always a good idea to make a figure look nice, that is, compensate leaps and awkward looking artifacts and focus on a plausible grouping. Do not hestitate to ungroup and regroup, if you feel a segment does not seem to reflect what you hear.
The results you get from careful editing are usually better than those that have been recognized automatically.
Di., 01.12.2009 - 00:38 Permalink
From my experience the anchor plays a very big role.
Two figure that looks the same but with different symbols as anchors will sound very different. Understanding how and why takes a fair bit of manual analysis.
As regrouping changes the anchors around, make sure you group before you do the figure recognition.
That said I don't feel at all confident in capturing an idea of interest for reuse, with the intent on maintaining its character.
I do find reusable phrases more useful as a raw medium with no expectations of the original beyond the rhythm. These have been useful in quickly auditioning ideas for accompaniment.