Skip to main content

Module 1 • Lesson 3🔢 32-bit, HDR & Blend-Mode Math

Eight bits is a delivery format; mastery works in more. 32-bit float is a scene-referred container that holds light beyond white, and once you know the actual arithmetic behind Multiply, Screen, and Add, blend modes stop being trial-and-error and become tools you aim.

📚 What You'll Learn

By the end of this lesson, you will be able to:

  • Explain bit depth as precision: 8-bit vs 16-bit vs 32-bit float
  • Understand 32-bit float as unbounded, scene-referred light (values above 1.0)
  • Merge a bracket to HDR and tone-map it back for an SDR display
  • Predict what Multiply, Screen, and Add/Linear Dodge actually compute
  • Use those modes deliberately for shadows, glows, and additive light

⏱️ Estimated Time: 55 minutes

🎯 Project (module deliverable): An HDR merge worked in 32-bit and tone-mapped to a clean SDR master, plus a glow built with additive blending — with the pipeline documented.

In This Lesson

👤 The Goal: Hold All the Light

A high-contrast interior, two ways. On the right, a single 8-bit exposure: the window blows to pure white and the corners crush to black — the dynamic range didn't fit. On the left, a 32-bit HDR merge holds the view through the window and the shadow detail, then tone-maps both into one frame. Drag the handle:

32-bit HDR Single 8-bit
A diagram of the principle. The scene's dynamic range is larger than one 8-bit exposure can hold; 32-bit float holds it all, and tone-mapping decides how to show it on a limited display.

🧠 Mental Model: Precision & the Unbounded 1.0

Bit depth is precision — how many distinct values a channel can hold. 8-bit gives 256 steps (fine for delivery, tight for heavy grading). 16-bit gives ~32,768 (comfortable editing headroom). 32-bit float is different in kind: values are floating-point and unbounded, so "white" at 1.0 is not a ceiling — a sky can be 8.0, a specular glint 60.0. That is a scene-referred container (Lesson 1.1) you can actually store.

This is why 32-bit and HDR go together. A single capture can't record the range of a scene with a bright window and a dark room, so you bracket (several exposures) and merge them into one 32-bit file that holds the whole range. Then tone-mapping compresses that range into what an 8-bit display can show — your creative choice, made last.

8 / 16-bit — bounded 1.0 = white (ceiling) anything brighter clips 32-bit float — unbounded 2.08.060.0 real highlight light lives above 1.0 — and survives
Figure 1: In 8/16-bit, 1.0 is a hard ceiling and anything brighter clips. In 32-bit float, values run past 1.0, so the sun and speculars keep their true relative brightness for grading and compositing.

🌗 HDR Merge & 32-bit

File ▸ Automate ▸ Merge to HDR Pro aligns a bracket and fuses it into 32-bit. Set the mode to 32 Bit and keep the toning slider neutral to preserve the raw range, then click OK to land a true 32-bit document. From there, Image ▸ Mode ▸ 16 Bits/Channel opens the HDR Toning dialog, where you compress the range to a display — this is the creative tone-map.

Merge to HDR Pro → 32-bit
−2 EV 0 EV +2 EV Mode 32 Bit ▾ ✓Remove ghosts 32-bit preview exposure (view only) then Image ▸ Mode ▸ 16-bit → HDR Toning to tone-map
Figure 2: Merge the bracket in 32 Bit mode to keep the full range. The preview exposure only changes what you see; the data stays intact until you tone-map down to 16-bit for delivery.

⚠️ Not everything works in 32-bit

Many filters and adjustments are unavailable in 32-bit float because they assume bounded 0–1 data. Do your range-critical work (merge, exposure, additive composites) in 32-bit, then convert to 16-bit for the bulk of your normal editing. 32-bit is a stage, not usually the whole session.

➗ What Blend Modes Compute

Blend modes are just per-channel math on the base (b) and blend (s) values, normalized 0–1. Learn three and you understand most of the panel:

Multiply r = b × s always darkens · shadows Screen 1−(1−b)(1−s) always lightens · glows Linear Dodge (Add) r = b + s true light sum · flares Why it matters • Multiply by 0 → black; by 1 → unchanged. Screen by 1 → white; by 0 → unchanged. • Add is real, additive light — the physically correct way to combine glows and flares… • …and it needs linear light (1.1) and 32-bit headroom (this lesson) to sum past 1.0 without clipping.
Figure 3: The three that unlock the rest. Contrast modes (Overlay, Soft Light) are just combinations of Multiply and Screen around mid-grey. Knowing the arithmetic lets you predict a result instead of cycling modes hoping.

✅ Add is where the last three lessons meet

Additive light (Linear Dodge) only behaves correctly when values are linear (Lesson 1.1) and can exceed 1.0 without clipping (32-bit, this lesson). A glow summed in gamma-encoded 8-bit clips and greys; summed in linear 32-bit it blooms like real light. That's the whole module in one operation.

🛠️ Guided Build: HDR + Additive Glow

Gather an exposure bracket (−2 / 0 / +2) of a high-range scene, or any image plus a bright element to glow.

Step 1: Merge to 32-bit · 10 min

  1. Merge to HDR Pro, set Mode 32 Bit, Remove Ghosts if needed, OK.
  2. Scrub the preview exposure to confirm both the highlights and shadows carry real detail.

Step 2: Tone-map down · 10 min

  1. Image ▸ Mode ▸ 16 Bits/Channel to open HDR Toning.
  2. Use Local Adaptation; compress the range with restraint (avoid the crunchy, haloed HDR cliché). This is your creative tone-map.

Step 3: Build an additive glow · 10 min

  1. On a bright source (a lamp, a sign), duplicate it to a layer, blur, and set Linear Dodge (Add).
  2. Do it once in 8-bit gamma and once in 32-bit/linear and compare — the linear glow blooms; the gamma one greys and clips.

Step 4: Deliver · 6 min

  1. Finish normal edits in 16-bit, then Convert to sRGB 8-bit for delivery (Module 5 covers output in full).
  2. Document the pipeline: bracket → 32-bit merge → tone-map → glow → deliver. Module 1 deliverable complete. 🏆

✅ Project Completion Checklist (Module 1 deliverable)

  • ☐ A bracket merged to a true 32-bit float file
  • ☐ Tone-mapped to 16-bit with restraint (no HDR crunch)
  • ☐ An additive glow built, compared gamma vs linear/32-bit
  • ☐ Delivered as sRGB 8-bit from a wide, high-bit master
  • ☐ The whole color pipeline written down and repeatable

🧗 Now You: Solo Variation

🌟 Your challenge

  1. Predict, then verify: sample two values and hand-compute Multiply, Screen, and Add before checking the Info panel. Trust the math.
  2. Build the same lens flare in 8-bit and in 32-bit linear; keep both as a permanent reminder of why the container matters.
  3. Tone-map one HDR three ways — natural, editorial, and deliberately overcooked — to calibrate your own taste for the line.

Going further: open a 32-bit EXR from a 3D render (you'll meet these in Module 2.3) and notice its values run far past 1.0 — the same unbounded, scene-referred light, straight from a renderer.

🍳 Recipe Card: 32-bit & Blend Math

Hold the range, aim the modes

  1. Bit depth = precision; 32-bit float = unbounded, scene-referred light
  2. Merge to HDR Pro (32 Bit) → tone-map to 16-bit with restraint
  3. Multiply = b×s (darken) · Screen = 1−(1−b)(1−s) (lighten) · Add = b+s (light sum)
  4. Do additive glows/flares in linear 32-bit so they bloom, not clip
  5. Range-critical work in 32-bit; convert to 16-bit for normal editing

Mantra: keep the light, and know the arithmetic that combines it.

📓 Learning Journal

Add to your journal after this lesson:

  • Key concepts you learned
  • Techniques that clicked for you
  • Questions or confusion points to revisit
  • Ideas you want to try
  • Your progress and feelings about learning this

✍️ This lesson's prompt: Which blend mode were you using "by feel" that you now understand as an equation? Write its formula in your own words.

📝 Module 1 Summary

🎓 Key Takeaways

  • The numbers in a file are a gamma-encoded record of light, not the light — and some edits belong in linear.
  • Scene-referred keeps unbounded light until a final display transform; display-referred clips early.
  • Camera Raw is the develop stage: profile and white balance first, tone from the ends, presence last, kept re-editable.
  • 32-bit float holds HDR light past 1.0; blend modes are math you can predict and aim.

🎉 What You've Accomplished — Across All of Module 1

You now understand color the way a colorist or compositor does: what the numbers mean, how light is encoded and decoded, how a raw becomes a master, and how bit depth and blend math govern every operation. This is the technical bedrock the rest of the course stands on — everything from photoreal compositing to prepress assumes it.

❓ Common Questions at This Stage

Should I just work in 32-bit all the time to be safe?

No — too many tools are disabled and files are heavy. Use 32-bit for the range-critical stage (merge, exposure, additive light), then convert to 16-bit for the bulk of editing and 8-bit only at delivery.

My HDR looks fake and crunchy. What went wrong?

Over-aggressive tone-mapping — too much local contrast and detail, lifted blacks. Pull Local Adaptation back toward a natural compression; the goal is to fit the range onto the display, not to advertise that you used HDR.

🔭 Looking Ahead

With color mastered, we make things disappear into a photograph. Module 2 — Photoreal Compositing uses everything here — linear light, 32-bit, blend math — to integrate an element so completely that no one can tell it was added.

✅ Before the Next Module

  • Develop one raw to a documented, re-editable master.
  • Merge one bracket to 32-bit and tone-map it cleanly.
  • Write your Learning Journal entry.

📚 Additional Resources

🌟 Encouragement for the Journey

You just went under the hood of color itself — the layer most Photoshop users never see. It's dense, and you got through it. Everything from here is more visual and more fun, and it all rests on what you now understand. On to making the invisible edit.