Hedra AI Generative media research, tested in the open.

Hedra AI

Generative media research, tested in the open.

Latest Articles

Pixel Borders, Real Budgets: Testing Ten Image Editors Where the Mask Ends
Experiments

Pixel Borders, Real Budgets: Testing Ten Image Editors Where the Mask Ends

We ran the same masked product-photo edits through ten image-editing APIs and graded the one thing users notice first: the seam. Here is the boundary-first protocol, the scorecard, and the production trade-offs that actually matter in 2026.

Your TTS Sounds Great Until the Feed Normalizes It: A Twelve-API Loudness Audit
Experiments

Your TTS Sounds Great Until the Feed Normalizes It: A Twelve-API Loudness Audit

We ran twelve text-to-speech and music-generation APIs through the same podcast delivery pipeline and measured integrated loudness, true peak, noise floor, and stereo behavior before and after platform normalization. Half of them sounded fine on laptop speakers and still failed spec. Here is the rig, the numbers, and the pre-flight check we now run before anything ships.

Aspect Ratio Is a Contract: Auditing Ten Image APIs for True Vertical Output
Experiments

Aspect Ratio Is a Contract: Auditing Ten Image APIs for True Vertical Output

We asked ten hosted image APIs for the same 9:16 vertical frame and then checked the actual pixels, subject placement, cropping behavior, and metadata. The results make a simple case: vertical output is a production contract, not a prompt suggestion.

Seven Ways to Say No: What Seven Media APIs Actually Do When Your Prompt Trips a Filter
Experiments

Seven Ways to Say No: What Seven Media APIs Actually Do When Your Prompt Trips a Filter

We ran the same 240 prompts through seven image, video, and voice APIs to see how each one handles a request it doesn't like. The refusals came in five distinct flavors, and only one of them actually told us the truth. Here's the taxonomy, the false positives we hit, and a test harness you can borrow.

Your Endpoint Is Not a Constant: Thirty Days of Watching Hosted Media APIs Shift Under Us
Experiments

Your Endpoint Is Not a Constant: Thirty Days of Watching Hosted Media APIs Shift Under Us

We pointed the same twelve prompts at ten hosted image and video APIs every morning for a month. The outputs changed while the URL, the docs, and the price stayed the same. Here is the capture routine we used, the signals that actually moved, and the archive you should build before the next silent rollout lands.

Alt Text Is Not a Vibe Check: Auditing What Generative Models Say About Their Own Output
Experiments

Alt Text Is Not a Vibe Check: Auditing What Generative Models Say About Their Own Output

Generative tools will happily describe their own images, videos, and audio, but those descriptions are often confident guesses dressed up as captions. We ran a small 2026 audit against human reviewers and came away with a practical workflow for accessible synthetic media that does not lie to screen readers.

The Half-Inch Problem: Six Image Models vs. the Ground Plane
Experiments

The Half-Inch Problem: Six Image Models vs. the Ground Plane

A product can be beautifully lit and still look fake if it hovers. We ran a fixed tabletop brief through six image generators, scored how convincingly each one put objects on a surface, and turned the failure modes into fixes creators can use before retouching.

The Invoice After the Invoice: What Generated Media Really Costs in Production
Experiments

The Invoice After the Invoice: What Generated Media Really Costs in Production

Vendors quote per image or per second. Production quotes per approved asset—after retries, review, storage, and delivery. Here is a reproducible ledger for pricing generative media work in 2026.

The Weights Are Not the Recipe: Reproducing an Open-Weights Generative Media Paper
Experiments

The Weights Are Not the Recipe: Reproducing an Open-Weights Generative Media Paper

A recent generative media paper shipped weights but no reference code, so we tried to rebuild the pipeline from scratch. Here is what broke, how close we got, and the checklist we now use before calling a release reproducible.

Milliseconds Are a Product Decision: Drafting a Real Latency Budget for Generative Media
Product Design

Milliseconds Are a Product Decision: Drafting a Real Latency Budget for Generative Media

Real-time generative media is less a model benchmark than a scheduling problem. We benchmarked an avatar and image pipeline across cloud, edge, and on-device deployments and turned the results into a latency budget worksheet product teams can adapt.

Fifty Shots With the Same Fake Bird: A Mascot Consistency Field Test
Product Design

Fifty Shots With the Same Fake Bird: A Mascot Consistency Field Test

We invented a US outdoor-brand mascot, built a reference kit, and pushed her through 50 stills and short clips. Faces drifted, jackets changed color, and one video made her blink. Here is the workflow that survived.

Ten Keeper Assets, Eight Tools, and a Spreadsheet Full of 'Almost'
Experiments

Ten Keeper Assets, Eight Tools, and a Spreadsheet Full of 'Almost'

We ran one 10-asset brief through eight 2026 generative media tools and tracked every retry, credit, licensing caveat, and edit minute. The sticker price was rarely the real number; cost per usable asset was.

The Webcam Is the Boss: Stress-Testing Talking-Head APIs in 2026
Experiments

The Webcam Is the Boss: Stress-Testing Talking-Head APIs in 2026

Studio demos make every talking-head API look like magic. We ran five of them through real US creator footage — regional accents, kitchen lighting, compressed webcam video — and scored sync, artifacts, latency, and cost. Here is the decision matrix we would actually use.

The Files Remember: Auditing Provenance on Generated Media Before You Publish
Product Design

The Files Remember: Auditing Provenance on Generated Media Before You Publish

Your export is carrying more than pixels. We audited the metadata, watermarks, and C2PA credentials left by common generative image, video, and audio tools, then traced what survives editing, screenshots, and social uploads. Here is the pre-publish checklist and the disclosure decisions product teams should make explicit.

Your Prompt Didn't Change—Your Pipeline Did: Instrumenting Generative Media for Reproducibility
Experiments

Your Prompt Didn't Change—Your Pipeline Did: Instrumenting Generative Media for Reproducibility

The same prompt can produce a different video three weeks later because the model, safety layer, codec, or default settings moved under your feet. Treat generative media like a dependency: log the run, hash the output, and regression-test the weird stuff.

The Voice Clone-Off: Pressure-Testing Speaker APIs Before Real Users Hear Them
Product Design

The Voice Clone-Off: Pressure-Testing Speaker APIs Before Real Users Hear Them

A cloned voice can pass a 15-second demo and still fall apart in a support call, a noisy car, or a 20-minute narration. Here is a compact 2026 test plan for speaker similarity, artifacts, latency, and cost—plus a procurement checklist that survives contact with real users.

Consent Is a Feature, Not a Footer: Seven UI Patterns Generative Media Teams Should Ship in 2026
Product Design

Consent Is a Feature, Not a Footer: Seven UI Patterns Generative Media Teams Should Ship in 2026

Policy pages do not stop a bad generation. These seven interface and API patterns, from authorization flows to consent receipts and C2PA export prompts, turn consent into something users can see, use, and revoke.

Ten Video Models, One 60-Second Brief: A 2026 Stress Test With Receipts
Experiments

Ten Video Models, One 60-Second Brief: A 2026 Stress Test With Receipts

We handed the same neo-noir short-film brief to ten generative video engines, capped the spend, and scored the outputs on consistency, camera control, audio, latency, cost, and licensing. Here is the prompt pack, the scoreboard, and the workflow guide we would actually use.