
A temporary stretch used for star removal is different from a finished creative stretch. Siril's documented legacy StarNet integration can stretch a linear image before processing and apply the inverse afterward. It does not establish that an arbitrary earlier custom stretch will be discovered and undone automatically.
Check the integration version before following old controls
The Siril 1.4.4 StarNet reference now carries a migration warning. Its legacy C interface supports older StarNet versions in Siril 1.4.x; StarNet 2.5 uses a Python script and is not supported by that old interface.
Identify the installed Siril and StarNet versions before using an integration-specific recipe. This article explains the documented transformation boundary. It does not give legacy console syntax as a universal command for the newer script.
Keep the original linear image available so a misunderstood stretch does not become the only processing source.
Identify which stretch was actually applied
For the documented legacy route, Pre-stretch linear image supplies an MTF stretch and its inverse around star removal. That is an explicit pair of operations.
If a custom MTF stretch has already been applied, the official integration tutorial says to record its low, mid and high parameters and use the corresponding inverse when further linear processing is required. Do not replace those parameters with a remembered visual impression.
A display that looks bright does not establish the stored image's processing history. Inspect the actual operations applied to the data.
Keep separated products in the same intended state
Retain the original, starless result and star component with clear identities. Do not combine a linear component with an independently stretched component merely because both look acceptable as standalone previews.
The reason for separating stars is to permit different treatments of the stars and remaining structure. The recomposition still needs an intentional processing relationship; star removal is not proof that all tiny bright features were classified correctly.
The RAW-development guide illustrates another boundary between recorded data and a rendered treatment, without defining these Siril operations.
Inspect reconstruction and artifacts honestly
Compare the recombined picture with the preserved input under a known treatment. Inspect bright stars, faint structure and places where the separation appears to remove real subject matter.
If banding appears, examine precision and transformations rather than assuming it is an astronomical feature; the gradient-banding guide addresses that diagnostic distinction.
No StarNet binary, Python integration, inverse transform or recomposition was run here. Check current script documentation for its controls; do not infer feature parity from this older interface's option names.
Related tasks
- Should I Develop a RAW Photo Before Using AI Image Editing?
- Why Does a Smooth Portrait Background Show Bands After Export?
Sources and scope
- Siril 1.4.4: StarNet Star Removal — deprecation and Pre-stretch: Legacy C interface remains for older StarNet in Siril 1.4.x but does not support StarNet 2.5, which uses a Python script. Legacy Pre-stretch applies MTF and its inverse; stars and background can be processed separately.
- Siril: Starnet++ integration — When to Remove Stars: For a custom prior MTF stretch, preserve lo/mid/hi parameters and apply invmtf manually if more linear processing is required. Tutorial describes older integration, not current 2.5 script options.
Official Siril documentation was opened and read on October 23, 2026 (Asia/Shanghai). Versioned references identify 1.4.4; legacy StarNet controls are explicitly distinguished from the newer 2.5 Python integration. Acquisition, registration, processing and acceptance checks are proposed, not performed. The cover is decorative.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: 85 LENS (Ngô Hoàng Auth) / Pexels, Pexels License. This is an illustrative source photograph, not an application screenshot or a generated test result.