Observed September 23, 2026 · Audited October 14, 2026
Made and Higgsfield: a matched-output cost comparison
Eighteen completed generations matched the named model, exact prompt, source image and native output specifications. Made debited $5.88 from its then-current wallet. Higgsfield debited 123 credits, equivalent to $5.945 when the entire $29 subscription is allocated to 600 fully used credits. This batch does not establish a substantial overall price advantage for Made.
These are historical account charges. Made funding fees are excluded; other Higgsfield subscription benefits are unvalued. The review was unblinded and makes no quality ranking or cost-per-usable-output claim. For current Made prices, see the price sheet.
18 / 18
Jobs completed
9 / 9
Native specification pairs matched
3
Attempts per task and platform
0
Independent human quality ratings
Actual debits and the accounting assumption
Every row contains three attempts on each platform. Quoted prices and actual debits agreed on all 18 runs. Higgsfield's allocated dollars are calculated as credits × $29 ÷ 600; they were not an additional cash payment for this batch.
| Task | Made wallet debit | Higgsfield debit | Higgsfield allocation |
|---|---|---|---|
| Nine-view product grid | $0.48 | 9 credits | $0.435 |
| Poster headline correction | $0.36 | 6 credits | $0.290 |
| Twelve-second floating video | $5.04 | 108 credits | $5.220 |
| All nine generations per platform | $5.88 | 123 credits | $5.945 |
The Higgsfield account had a paid $29 invoice dated August 30, 2026 and a 600-credit monthly allowance. Its usage ledger recorded three debits each of 3, 2 and 36 credits. The dashboard separately displayed $4.92 for those 123 credits; that display is a different allocation from the invoice-based $5.945 used above.
Fully using the plan is an assumption. Using only half of its 600 credits would double the amount allocated to each used credit, if the entire invoice were still assigned to credits. The value of other plan benefits and Made's historical funding fees is unknown here.
What was matched
Nine-view product grid
Nano Banana 2 · 4K · 3:4
3584 × 4800 native pixels in every downloaded image
Poster headline correction
Nano Banana 2 · 2K · 3:4
1792 × 2400 native pixels in every downloaded image
Twelve-second floating video
Wan 3.0 Prime, reference mode · 720p · 9:16
720 × 1280 native pixels · 12.000 seconds · 360 frames · 30 fps · audio present
The protocol and input hashes were saved before submission. Every attempt used the same exact prompt and source bytes, without tuning or replacement runs. No scaling or retiming was used to qualify the original outputs as a match. Hidden model builds, preprocessing and identical random seeds were not verified. These were general-generator tests, separate from proprietary Shots, automatic text detection and Floating Fall workflows.
Every attempt, input and exact prompt
All 18 outputs are below, including imperfect results. Images are scaled to 1,200 pixels wide; videos are 360 × 640 review copies with the original duration, frame count and audio. These compressed copies are for inspection; native dimensions and original SHA-256 hashes are in the dataset. The larger original downloads remain archived privately.
Nine-view product grid
Exact frozen prompt
Create a cinematic contact sheet from the reference image: exactly 9 photographs in an edge-to-edge 3 by 3 grid. Each cell must occupy exactly one third of the total width and one third of the total height. No gutters, borders, margins, text, numbers, captions or watermarks. Show the SAME subject and moment from nine distinct camera positions. Preserve the reference subject's identity, design, materials, wardrobe, surroundings, lighting and photographic style. Do not turn it into a drawing. Change only the camera position, framing and focal length. Do not invent a narrative or different subjects. Row 1: wide establishing view; medium view; close-up. Row 2: low angle; high angle; profile side view. Row 3: three-quarter view; reverse angle; extreme detail. All nine cells must be fully filled photographs with clean straight divisions at exactly 1/3 and 2/3 of the image. The overall sheet is 3:4 and every cell is 3:4.
Poster headline correction
Exact frozen prompt
Edit the supplied poster image. Replace ONLY the large headline "CREATE WITH INTEND." with "CREATE WITH INTENT.". Keep the original line breaks and placement. Preserve the composition, origami shapes, blue and pink background, colors, lighting and all other existing text. Match the original white typography, font weight, spacing and texture. Fit the complete poster into a 3:4 portrait image without removing any existing content. Do not add captions, borders, watermarks, or a before/after layout. Return one complete edited image.
Twelve-second floating video
Exact frozen prompt
Create a 12-second photorealistic video using Image 1 as the sole visual reference. Preserve the same man, face, rust overshirt, white T-shirt, green trousers, coffee cup, cafe interior and lighting. Use one continuous shot with no cuts. For the first 3 seconds he stands naturally holding the cup. From 3 to 9 seconds gravity gently slows and he leans backward into a controlled, graceful horizontal float; his clothing and coffee cup follow believable slow motion. The camera smoothly tracks him and keeps his whole body visible. From 9 to 12 seconds he remains safely suspended above the cafe floor. Keep the same location throughout. No collision, ground impact, injury, extra people, inserted products, captions, logos, dialogue or music. Include quiet cafe ambience.
Evidence and limits
On October 14, 2026 we rechecked the hashes and native specifications of all 18 archived output files, and re-read Made's nine original job receipts through its read-only API. All nine jobs still reported success and charges totaling 588 cents. Higgsfield billing evidence is a contemporaneous transcription of its signed-in ledger, plan and paid invoice pages. Those private pages are not part of the public data.
- One source per task and three attempts per platform. This small convenience sample cannot establish a general price, failure-rate or quality advantage.
- These are dated account charges, not current prices. Made used its then-current wallet. Higgsfield discounts and the specific paid plan affect the comparison.
- The invoice allocation assumes all 600 credits are used and assigns no value to other subscription benefits. Made historical funding fees are excluded.
- The review was unblinded and contains no independent human quality ratings, cost-per-usable-output result, latency comparison or statistical quality ranking.
- Matching named models and visible settings does not verify hidden provider build IDs, preprocessing or random seeds.
- The general generators were tested. Proprietary native tool pipelines and the earlier unmatched pilot are excluded.
- Public media are scaled review copies. Native metadata and SHA-256 hashes refer to the preserved original downloads; private billing pages and original output files are not included in the public dataset.
Made publishes this comparison of its own service. It is not an independent evaluation. No new generations were commissioned for this report.
The JSON dataset contains all 18 receipt summaries, original and review-file hashes, measured dimensions, prompts, accounting assumptions and source-archive fingerprints. This observational report is separate from the preregistered cost-per-usable-output study, which still requires complete runs and two blinded human raters.