Face Shape Detector: Compare Methods, Free Access and Exports
Choose a face shape detector by its disclosed method, accepted photo format and result access—not by a confidence score alone; this comparison records what six official tool pages disclose and leaves undocumented limits unresolved.
Download the comparison table (CSV)
Collected 2026-09-22 (UTC). This is a dated audit of official pages, not a test of detection accuracy. No portraits were submitted, quotas exhausted or result files exported. “Not stated” means not found on the linked page during this visit; it does not mean the feature is absent. Camera and upload controls were inspected without activating a camera. The listed processing methods and privacy descriptions are provider claims, not code-audit findings.
| Tool | Detection method | Input | Output | Login | Free allowance | Accuracy claim | Export | Other limits |
|---|---|---|---|---|---|---|---|---|
| AltFTool official page 2026-09-22 (UTC) | 68 landmarks; face-api.js and geometric ratio thresholds | Image file or drag-and-drop; extensions not stated; webcam explicitly excluded | Shape, confidence figure and proportion readouts | No account required (stated) | Free; no usage limit (stated) | Calls it a measured estimate; no validated accuracy percentage stated | Not stated | File-size and resolution limits not stated |
| KnowYourFaceShape official page 2026-09-22 (UTC) | MediaPipe; 478 landmarks; proportion matching | Camera or file; JPG, PNG, WebP, HEIC | Shape, measurements, ratios and confidence | No account required (stated) | Free; numerical usage allowance not stated | Explicitly declines to publish an independently unverifiable accuracy percentage | Not stated | File-size, resolution and daily limits not stated |
| Symmi official page 2026-09-22 (UTC) | 478 landmarks; widths, face length and jaw angle compared with shape ranges | File, drag-and-drop or paste; JPG, PNG, WebP | Shape breakdown and measurement ratios | Not stated; upload control visible without signing in | Free; numerical usage allowance not stated | Describes measurement-based results; validated accuracy percentage not stated | Not stated | File-size, resolution and daily limits not stated |
| Style Look Lab official page 2026-09-22 (UTC) | MediaPipe plus proprietary classifier and normalized 36-point contour | Photo file; extensions not stated | Closest two of seven labels, with checks and supporting measurements | No sign-up (stated) | Free; numerical usage allowance not stated | Explicitly says its internal rules have not been independently validated | Not stated | File-size, resolution and daily limits not stated |
| Facial Recognition official page 2026-09-22 (UTC) | 68 landmarks; ratio distances; estimated hairline-to-chin length | File or drag-and-drop; JPG, PNG, WebP recommended | Shape percentage breakdown and ratios | No account required (stated) | Free; as many photos as desired (stated) | Claims greater consistency than mirror judgement; no validated accuracy percentage stated | Not stated | Largest face analysed in a group photo (stated); file-size and resolution limits not stated |
| DetectMyFaceShape official page 2026-09-22 (UTC) | MediaPipe; 478 landmarks; ratios scored against shape profiles | Camera, file or drag-and-drop; JPG, PNG, WebP | Shape, confidence score, measurements and styling tips | No sign-up (stated) | Free; no usage limit (stated) | Uses accuracy and precision language; no validated accuracy percentage stated | Downloadable result card stated; file extension not stated | File-size and resolution limits not stated |
What this comparison lets you distinguish
AltFTool, Facial Recognition and DetectMyFaceShape explicitly allow unrestricted repeat use in the pages checked. KnowYourFaceShape, Symmi and Style Look Lab describe free access but do not specify a numerical allowance there. Free access and an explicit promise of unlimited use therefore occupy different cells in this record.
KnowYourFaceShape lists HEIC beside its photo input. Symmi, Facial Recognition and DetectMyFaceShape name JPG, PNG and WebP. AltFTool and Style Look Lab do not enumerate extensions in the inspected text. That is a documentation difference, not proof that an unlisted format will fail.
DetectMyFaceShape advertises a downloadable result card. The inspected page does not identify its extension. The remaining inspected pages do not specify an export format. A result shown in a browser is not evidence of a PDF, CSV or image export; this comparison file is our own downloadable research table, not an export produced by those detectors.
Do not rank these tools by their displayed percentages
KnowYourFaceShape declines to publish an unverifiable accuracy percentage, while Style Look Lab says its internal rules have not been independently validated. AltFTool calls its result an estimate. These statements disclose limits; they do not establish which tool makes fewer classification errors.
Facial Recognition describes percentage scores derived from distances to shape ratios. DetectMyFaceShape calls its result a confidence score. Neither description supplies a shared evaluation protocol that would let this audit compare their percentages as accuracy measurements. We have therefore recorded the claim, rather than inventing a performance ranking.
The landmark counts in the table identify the methods described by their providers. They are not successful-test counts. More landmarks alone cannot establish a winner from this evidence. This audit contains no labelled test set and makes no accuracy calculation.
A file you can revisit, not a permanent specification
Keep the source URL and collection date attached to any row you reuse. If a provider changes its upload screen, allowance or export option, update that field from the new page and retain the earlier observation separately. Do not silently replace an unknown limit with a guessed value.
For a workflow that requires a particular file format or a saved report, an unresolved cell is a reason to check that requirement before relying on the tool. The useful next question is precise: does this page accept the file I have, and does its actual result offer the output I need? The current record answers only what the official pages disclose.
Dated source record
Each line below refers to the linked official page observed on the stated collection date. Specifications are reported claims, not measured performance.
- AltFTool official page — 2026-09-22 (UTC); About / How to use / FAQ. Detection method: 68 landmarks; face-api.js and geometric ratio thresholds. Input: Image file or drag-and-drop; extensions not stated; webcam explicitly excluded. Output: Shape, confidence figure and proportion readouts. Login: No account required (stated). Free allowance: Free; no usage limit (stated). Accuracy claim: Calls it a measured estimate; no validated accuracy percentage stated. Export: Not stated. Other limits: File-size and resolution limits not stated.
- KnowYourFaceShape official page — 2026-09-22 (UTC); Photo input / How it works / FAQ. Detection method: MediaPipe; 478 landmarks; proportion matching. Input: Camera or file; JPG, PNG, WebP, HEIC. Output: Shape, measurements, ratios and confidence. Login: No account required (stated). Free allowance: Free; numerical usage allowance not stated. Accuracy claim: Explicitly declines to publish an independently unverifiable accuracy percentage. Export: Not stated. Other limits: File-size, resolution and daily limits not stated.
- Symmi official page — 2026-09-22 (UTC); Upload area / How it is decided. Detection method: 478 landmarks; widths, face length and jaw angle compared with shape ranges. Input: File, drag-and-drop or paste; JPG, PNG, WebP. Output: Shape breakdown and measurement ratios. Login: Not stated; upload control visible without signing in. Free allowance: Free; numerical usage allowance not stated. Accuracy claim: Describes measurement-based results; validated accuracy percentage not stated. Export: Not stated. Other limits: File-size, resolution and daily limits not stated.
- Style Look Lab official page — 2026-09-22 (UTC); Tool / Sources & limitations. Detection method: MediaPipe plus proprietary classifier and normalized 36-point contour. Input: Photo file; extensions not stated. Output: Closest two of seven labels, with checks and supporting measurements. Login: No sign-up (stated). Free allowance: Free; numerical usage allowance not stated. Accuracy claim: Explicitly says its internal rules have not been independently validated. Export: Not stated. Other limits: File-size, resolution and daily limits not stated.
- Facial Recognition official page — 2026-09-22 (UTC); Upload area / Method / FAQ. Detection method: 68 landmarks; ratio distances; estimated hairline-to-chin length. Input: File or drag-and-drop; JPG, PNG, WebP recommended. Output: Shape percentage breakdown and ratios. Login: No account required (stated). Free allowance: Free; as many photos as desired (stated). Accuracy claim: Claims greater consistency than mirror judgement; no validated accuracy percentage stated. Export: Not stated. Other limits: Largest face analysed in a group photo (stated); file-size and resolution limits not stated.
- DetectMyFaceShape official page — 2026-09-22 (UTC); Upload Photo tab / How to use / Features / Method. Detection method: MediaPipe; 478 landmarks; ratios scored against shape profiles. Input: Camera, file or drag-and-drop; JPG, PNG, WebP. Output: Shape, confidence score, measurements and styling tips. Login: No sign-up (stated). Free allowance: Free; no usage limit (stated). Accuracy claim: Uses accuracy and precision language; no validated accuracy percentage stated. Export: Downloadable result card stated; file extension not stated. Other limits: File-size and resolution limits not stated.
What remains unverified
Actual file decoding, successful detection, accuracy, server-side image handling, enforced quotas and exported file types were not tested. We do not turn undocumented fields into “no,” and we do not claim this is the only comparison on the English-language web.