How to estimate high/low energy segments in Spotify tracks without preview_url?

[Kiddo1337]

Helluuur!

I'm working on a personal project that involves analyzing the energy of Spotify tracks and adapting playback accordingly.

In many cases, the tracks lack a preview_url, which makes traditional audio analysis impossible. To work around this, I'm using metadata (title, artist, duration) to infer approximate energy levels and then seek to parts of the track via:

• GET /v1/me/player/currently-playing → Get current track

• PUT /v1/me/player/seek → Jump to specific timestamp

My current fallback approach:

• Estimate song structure (intro, verse, chorus, etc) using duration-based percentages.

• Use energy-related keywords (e.g. "remix", "chill") to influence the estimated energy.

• Jump to “high energy” timestamps (e.g. 60–80% in) or “low energy” (e.g. 5–20%).

The challenge:

• Some tracks don't respond well to seek, or jump back.

• I want to ensure that I'm not violating any terms by segmenting tracks based on inferred or API-fetched data.

• I’d like to improve accuracy of this logic – ideally without using preview_url or actual audio playback.

Questions:

1. Are there recommended practices for estimating song energy or structure without preview_url?

2. Are audio-features or audio-analysis endpoints useful for this – and can I cache those results?

3. Are there technical limitations on how PUT /seek behaves with certain tracks?

4. Any known tools or best practices for approximating segments (chorus vs verse) via metadata?

Would really appreciate any pointers or thoughts! 
Thanks in advance,
– Fares