Reference G · Still open

Yasser Ashour

12 minutes ago

24 min read

Reference

A question is open when no maker, no measurement and no finished job has answered it. This reference gathers every open question in the chapters, once, in production order, with what the book teaches meanwhile and what would settle it.

In this reference

Every open question in the book, gathered from the boxes at the end of the chapters and from each tool's version notes, and merged where two chapters ask the same thing
For each: what the book teaches until it is settled, and the smallest thing that would settle it
Where it bears on the teaching, and the sections whose boxes carry it

How to read an entry. The chapters' boxes hold 129 open questions between them; where two chapters ask the same thing, they are merged here into 104 entries, in ten groups that follow the production. A question is listed here because the book could not settle it from a maker's page, a measurement or finished work; the chapter that depends on it teaches the cautious rule in the column beside it. Where the answer is one cheap test, the test is named; where it is a person's decision or a written answer, the person is named. A question a source has since answered is not listed. Re-read a tool's version notes before a client job, because the pages this list rests on are dated.

G.1 Concept, treatment and planning

Chapters 3 and 4.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
Regional practice in making ideas. Public accounts of how ideas are made in Cairo, Riyadh and Dubai are thin.	Treat each pattern in the concept chapter as a test to run past a native reader, not as a verdict on the local audience.	A native reader's judgement of each pattern for the market in hand, or fuller public accounts of regional practice.	Bears on 3.2 What remembered work has in common. Listed in 3.9.
Whether a score can choose between ideas better than a written comparison.	Compare ideas with reasons, and name a recommendation and the strongest alternative.	The same ideas both scored and argued, with the director judging which method chose better.	Bears on 3.7 Testing an idea. Listed in 3.9.
Timing for Egyptian and Gulf dialects. No robust open figures exist.	Time every line by native reads, and treat the locked take as the authority.	A published timing figure for each dialect, or the project's own locked takes, timed.	Bears on 4.5 The first seconds, and how long a line takes. Listed in 4.9.
Platform guidance moves. The hook and text-density figures in 4.5 are the platform's guidance as of its June 2025 page.	Re-read the page for the platform you are delivering to before you write a hook.	The platform's current guidance, read on the day.	Bears on 4.5 The first seconds, and how long a line takes. Listed in 4.9.

G.2 Words, camera, light and the world

Chapters 2 and 5 to 7.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
How a model reads camera height and distance in real units, how it weighs a number against the words beside it, and how it weighs the order of the elements. No maker publishes any of them.	Write each number beside its visible result, and judge the returned frame, not the number.	A maker's statement, or a like-for-like comparison judged on the returned frames.	Bears on 2.5 Numbers, each beside its result; 5.7 Placement and lens; 9.3 Numbers beside results. Listed in 2.10, 5.11, 9.12, 11.10 (Nano Banana Pro), 12.10 (GPT Image 2.5).
Where in an action to freeze a keyframe. The 20 to 40 per cent band rests on a house rule and on finished practice, not on a maker's measurement.	Start in the band and adjust by what the clip does with the frame.	Freezes across the band on the same action, judged on the clips they make.	Bears on 9.4 The frozen instant, and the frame inventory. Listed in 2.10, 9.12.
Whether labelled sections beat one paragraph for a photographic frame on GPT Image 2.5.	Use one paragraph for a photographic frame, and labelled sections for layouts and text-heavy briefs.	The same frame written both ways on the same tool, judged by the director.	Bears on 2.4 Choose the writing format by the job. Listed in 2.10.
How each tool reads a Kelvin figure, as a colour, as a cue for warmth or not at all, and what a figure does in a black-and-white frame.	The rule rests on craft and finished practice, with a visible result beside every figure. Check the first returned frame.	A maker's statement, or one figure varied on a fixed frame on each tool.	Bears on 6.4 Colour temperature, and two light worlds. Listed in 6.10.
Contrast ratios by register. They are rules of thumb from published films, not a measured method.	Give one ratio with its reason, and expose for the skin first.	A measured method, or the director's calibration on returned frames.	Bears on 6.3 Contrast and exposure. Listed in 6.10.
Whether a prompted capture finish and a delivery grain laid in post work well together. It has not been tested.	Judge a representative frame before the look is locked.	One representative frame with the finish prompted and the grain laid over it, judged before the look is locked.	Bears on 6.6 The finish, in two layers. Listed in 6.10.
The realism check, the six tells and the material-families rule. They come from published craft and from studies of generated images; they are working checks, not measured methods.	Use them as checks on the returned frame.	A measured study of the checks on generated commercial work.	Bears on 7.1 Why a generated world looks built; 7.9 The check before you sign a frame. Listed in 7.9.
Whether each video tool re-renders props and materials from the start frame differently.	Protect the key props by name until a tool's own page says otherwise; the sections in 7.8 check each tool.	Each tool's own page, or the same start frame run on each tool and compared.	Bears on 7.8 Which tool for which job. Listed in 7.9.

G.3 Stills, identity and references

Chapters 8 to 12.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
Neutral master or in-look anchor for a film whose look may change, and which holds a face better over a long run. No maker and no comparison settles it.	A neutral master first, then the in-look anchor made from it. On a commercial with changing light, start from the neutral master.	The two routes compared over a long run on one film whose look is still open.	Bears on 8.8 The continuity plan, and the order of work; 10.2 Master, anchor and in-look anchor. Listed in 8.10, 10.14.
Whether a locked identity clause improves identity on top of the pictures.	Write the clause once from the approved master and paste it unchanged. The master wins any disagreement (10.4).	One run with the clause and one without, on the same references, judged against the master.	Bears on 10.4 The identity clause. Listed in 10.14.
Whether a host platform's image tool needs an operation phrase. Google warns that a chat model may answer in text unless asked to create an image.	If a tool ever returns text, add "Create a photograph of …" ahead of the medium.	The platform's documentation, or the first job that returns text.	Bears on 9.2 The twelve elements. Listed in 9.12.
Which sheet layout holds a character best in profile.	The 3-panel is the default; the 5- or 6-panel is the variant for a character seen in profile or turning.	The layouts run on the same character and compared in profile.	Bears on 10.6 The master's grammar, sheets and single views. Listed in 10.14.
Whether a real person's likeness holds on the release route.	Before a client job, make one frame from a consenting colleague's photographs and check it at 200 per cent.	That one frame, checked at 200 per cent, with the release in writing.	Bears on 10.8 Real people, and an actor's name as a rights decision. Listed in 10.14.
Whether a finished-grade edit holds a signed face.	Judge each edit against the parent.	A run of edits on one signed face, each compared with its parent.	Bears on 10.13 The one-change edit. Listed in 10.14.
Whether GPT Image 2.5 can carry a signed identity, and which of the two stills tools is the better camera for a photographic keyframe from text. No maker compares them.	Signed faces stay on Nano Banana Pro. The routing follows the documentation and the finished work.	A like-for-like comparison on finished-grade prompts, judged by the director.	Bears on 9.7 Which tool for which job; 12.1 What it is (GPT Image 2.5). Listed in 10.14, 12.10 (GPT Image 2.5).
What runs under the platform's "Nano Banana Pro". The job history records the label nano_banana_2 while the bill reads "Nano Banana Pro" (27 Sep 2026).	Log both labels on every job, and record "Nano Banana Pro route (recorded as nano_banana_2)".	Higgsfield's answer.	Bears on 11.1 What it is (Nano Banana Pro); 33.7 Checks before you sign (Higgsfield). Listed in 11.10 (Nano Banana Pro), 33.10 (Higgsfield).
How many pictures an image tool takes, and whether their order binds. Higgsfield publishes no reference ceiling for Nano Banana Pro; the same is open for GPT Image 2.5.	Plan on five pictures or fewer, and treat six or more as an experiment.	The platform's documentation, or one job with the order swapped.	Bears on 11.2 How it reads what you give it (Nano Banana Pro); 12.2 How it reads what you give it (GPT Image 2.5). Listed in 11.10 (Nano Banana Pro), 12.10 (GPT Image 2.5).
Which sizes the platform returns outside 2k at 16:9. That size has been measured (2752 × 1536); the others in Figure 11.1 are Google's table.	Read the size of the first file you make at any new setting.	One frame at each size, read from the file.	Bears on 11.6 Settings (Nano Banana Pro). Listed in 11.10 (Nano Banana Pro).
Whether Higgsfield's mask on Nano Banana 2 is worth a route. It has not been tried.	The fence in words and the composite in post are the routes.	One masked edit on the route, compared with the same edit fenced in words and composited in post.	Bears on 11.6 Settings (Nano Banana Pro). Listed in 11.10 (Nano Banana Pro).
What GPT Image 2.5 does with Arabic, and which file a transparent job returns on the host.	Read both on the returned file.	The first returned file of each.	Bears on 12.3 The prompt, in this tool's dialect (GPT Image 2.5); 10.12 Text in the frame: two layers. Listed in 12.10 (GPT Image 2.5).
Whether the host's Sunburst is exactly OpenAI's 8 Sep snapshot.	Treat OpenAI's guide as the guide to the host's Sunburst, and judge every returned frame.	Higgsfield's answer.	Bears on 12.1 What it is (GPT Image 2.5). Listed in 12.10 (GPT Image 2.5).
Sunburst against Flare on the edits a commercial needs.	Before a client job depends on Sunburst, run one edit on both with the same file and prompt and only the variant changed (about 5.5 credits at 27 Sep prices).	That paired edit.	Bears on 12.6 Settings (GPT Image 2.5). Listed in 12.10 (GPT Image 2.5).

G.4 Routes, the moving shot and performance

Chapters 13 to 15.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
Whether Kling 3.0 takes a prompt with no start frame, or an end frame alone, on the connection.	Give Kling a start frame.	A free price check with nothing attached, or with only an end frame, settles it.	Bears on 13.4 When to leave the start frame behind. Listed in 13.7, 16.10 (Kling).
Whether Seedance's edit and extension modes pin the task, what they cost on the connection, whether an extension returns the new part or the whole clip, and whether the source's sound is kept.	Set the mode, and keep other operations' task words out.	One cheap job of each, after a free price check.	Bears on 13.2 The operations. Listed in 13.7, 17.10 (Seedance 2.5).
Kling 4.0 (announced 28 Sep 2026). Which of its announced references and keyframes reach the connection, what it costs, how it handles Arabic, and whether 3.0 stays listed.	Teach and plan on Kling 3.0. Read the connection's list before anything is planned on 4.0.	The connection listing 4.0, then a free price check and one probe.	Bears on 13.6 Which tool for which operation. Listed in 13.7, 16.10 (Kling).
Whether Kling honours a stated distance, and how closely. No maker documents it. Kling 4.0 may change the answer.	Size a move by the measured-move method (14.5), and treat a stated distance as a cue whose result you check on the returned clip.	A maker's statement, or the measurement on your own clips repeated on 4.0.	Bears on 14.5 Measured moves: sizing one, and measuring it. Listed in 14.16, 16.10 (Kling).
Whether Seedance 2.5 and Wan 3.0 respond to an approximate distance in prose. Neither maker documents distances.	Their route is pacing words, ordered beats and whole-second stamps (17.3, 18.3). Use a distance only as a repair.	A maker's statement, or one clip each with a distance stated.	Bears on 17.3 The prompt, in this tool's dialect (Seedance 2.5); 18.3 The prompt, in this tool's dialect (Wan 3.0). Listed in 5.11, 14.16.
Whether the connection passes Seedance 2.5's whole-second stamps and sound notation, and Kling's in-take markers, or speaks or prints them.	Try one 480p job before a shot depends on either.	One run of each. It is a route fact.	Bears on 17.3 The prompt, in this tool's dialect (Seedance 2.5); 16.3 The prompt, in this tool's dialect (Kling). Listed in 14.16, 16.10 (Kling), 17.10 (Seedance 2.5).
Camera devices, reverse motion and the composite repair, and compound devices such as the dolly zoom. They are written from craft and from other directors' reports; none has been proved on a current model, and none has been run on Kling.	Test one on a cheap frame before it carries a scene.	One cheap run of each on the tool that will carry the shot.	Bears on 14.9 Camera devices; 14.13 Reverse motion, and the composite repair. Listed in 5.11, 14.16, 16.10 (Kling).
Which frame to export for the next start frame. The maker's return-last-frame option is not on the connection.	Export the accepted last frame by eye.	The connection offering the option.	Bears on 14.8 Start and end frames. Listed in 14.16.
Whether a bounded path keeps a turn inside three-quarter. Turns pinned with an end frame have not been run.	Treat it as untried, and do not let a client shot depend on it until it has been run.	One turn pinned with an end frame, run and judged.	Bears on 14.11 Turns. Listed in 14.16.
A recorded performance reference, sent to Seedance as a video reference, for subtle acting.	A lever worth a scratch test on the first project that needs it.	That scratch test.	Bears on 15.4 The task and the cues; 17.2 How it reads what you give it (Seedance 2.5). Listed in 15.10.
Two speakers, or two speaking faces, in one frame, on any route.	Prove it on a scratch clip before a client shot depends on it.	That scratch clip, run on each route.	Bears on 15.6 Listeners, handovers and two-shots; 22.3 Design the sync out first. Listed in 15.10, 22.9.
Wan 3.0, Veo 3.1 and MiniMax H3 as actors. There is no reliable reading yet.	Test on a scratch clip before a client shot depends on them.	That scratch clip, on each tool.	Bears on 15.9 Which tool for which job. Listed in 15.10.

G.5 The video tools

Chapters 16 to 19.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
What Kling's 4k mode is, native or upscaled, with its raster and frame rate, and what Seedance's 1080p returns.	Probe every file for raster, frame rate, frame count and length the first time a setting is used.	One probe of each.	Bears on 16.6 Settings (Kling). Listed in 16.10 (Kling), 33.10 (Higgsfield).
Whether fifteen-second clips hold on Kling. Never run.	Plan takes of five to eight seconds; ten seconds has run once and fifteen never.	One fifteen-second clip, run and judged.	Bears on 16.6 Settings (Kling). Listed in 16.10 (Kling).
What Omni Edit accepts and returns on the connector, and what it costs.	Treat it as untried, and do not let a client shot depend on it until it has been run.	A price check, then one job inside a packet with a cap.	Bears on 16.3 The prompt, in this tool's dialect (Kling). Listed in 16.10 (Kling).
Whether 9:16 and 1:1 clips are good on Kling. Three 9:16 clips ran without a verdict; 1:1 never ran.	Only 16:9 has been judged. Set the aspect every time, and make the keyframe at the delivery aspect first (14.15).	A 9:16 and a 1:1 clip, run and judged.	Bears on 14.15 Format, length, resolution, handles and probing; 16.6 Settings (Kling). Listed in 16.10 (Kling).
Which terms govern a clip made through the connector, Kling's or Higgsfield's.	Both sets apply. Read them again before a client job.	Higgsfield's or Kling's written statement of which governs.	Bears on 16.9 Limits, prices and rights (Kling). Listed in 16.10 (Kling).
Whether the connection keeps your upload order as the Image numbers on Seedance 2.5, and how a start frame and image references share numbers.	Name every file in words.	One 480p job with two images in swapped order.	Bears on 17.2 How it reads what you give it (Seedance 2.5). Listed in 17.10 (Seedance 2.5).
How closely a generated English voice on Seedance 2.5 follows a sample.	Run one 480p job with the actual sample and listen before a client shot depends on it.	That job.	Bears on 17.3 The prompt, in this tool's dialect (Seedance 2.5). Listed in 17.10 (Seedance 2.5).
A real actor's photograph as a Seedance reference. The maker refuses real faces; the connection accepted generated ones.	Plan real actors as unproven.	One 480p job with a real actor's photograph, run only with the actor's release in hand.	Bears on 10.8 Real people, and an actor's name as a rights decision; 17.2 How it reads what you give it (Seedance 2.5). Listed in 17.10 (Seedance 2.5).
Whether sound on returns your sample's own speech or your line in the sample's voice on Wan 3.0.	Run one 480p job with a sample on a different sentence and compare the track with the upload before a shot depends on it.	That job.	Bears on 18.3 The prompt, in this tool's dialect (Wan 3.0). Listed in 18.10 (Wan 3.0).
Whether the connection carries a clip to Wan 3.0 as a video reference, an edit or an extension, and what it costs.	Ask for the price at no cost with the clip attached, then run one 480p job.	The price check and that job.	Bears on 18.2 How it reads what you give it (Wan 3.0). Listed in 18.10 (Wan 3.0).
More than one image reference on Wan 3.0 (two people, or a person and a product), a line longer than about five seconds, and two speakers. None has been run.	Treat it as untried, and do not let a client shot depend on it until it has been run.	One 480p job of each.	Bears on 18.3 The prompt, in this tool's dialect (Wan 3.0). Listed in 18.10 (Wan 3.0).
Whether the preview label on Wan 3.0 lets limits, prices or output change without notice.	Re-read the pages before a client job.	The maker's statement of what the label means.	Bears on 18.1 What it is (Wan 3.0). Listed in 18.10 (Wan 3.0).
What Veo 3.1 returns at high quality against basic at 4 s, and whether the sound lands on the contact frame as written.	Treat it as untried, and do not let a client shot depend on it until it has been run.	A basic clip and a high clip on the same prompt, probed and heard.	Bears on 19.9 Limits, prices and rights (Veo 3.1 and MiniMax H3). Listed in 19.10 (Veo 3.1 and MiniMax H3).
Whether MiniMax H3's 2K is real detail, what its track contains, and whether an audio reference ever drives the lips.	Treat it as untried, and do not let a client shot depend on it until it has been run.	One H3 clip at 2K, probed and viewed at full size, with its track heard.	Bears on 19.9 Limits, prices and rights (Veo 3.1 and MiniMax H3). Listed in 19.10 (Veo 3.1 and MiniMax H3).
Whether a start frame and references can be attached together to a Veo or H3 job on the connection.	Treat it as untried, and do not let a client shot depend on it until it has been run.	A free price check with both attached.	Bears on 19.2 How it reads what you give it (Veo 3.1 and MiniMax H3). Listed in 19.10 (Veo 3.1 and MiniMax H3).
Whether Gemini Omni Flash 1.1 should replace Veo 3.1 as the sound-led route once its contract on the connection is read (Chapter 20).	Veo 3.1 keeps the sound-led role, to be checked on a scratch clip first (19.7).	The connection listing Omni, and its contract read.	Bears on 20.1 How to read an entry. Listed in 19.10 (Veo 3.1 and MiniMax H3).

G.6 Voices and the speaking shot

Chapters 21 to 24.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
Which model reads Egyptian, Najdi and Hejazi best, Eleven v4 or Eleven v3. No maker page says.	Audition both on every new Arabic voice, with a native ear, and write the model on every take.	The audition and a native ear, voice by voice.	Bears on 21.3 Three roads to a voice; 23.2 How it reads what you give it (ElevenLabs). Listed in 21.11, 23.10 (ElevenLabs).
Whether a repaired word survives a video model's new performance, and how Seedance 2.5 and Kling 4.0 handle Egyptian and Saudi Arabic on the new-performance route.	Only a native ear on the clip can say (22.7).	A native ear on the returned clip.	Bears on 22.6 Route B: a new performance; 21.7 The Egyptian line workflow: repair, measure, lock. Listed in 21.11, 22.9.
Whether a library voice may be used in an advertisement. The licence texts are silent or unclear.	The client's legal side decides in writing (21.6, 23.9). Ask before a final rests on one.	The client's legal side, in writing, or a clear statement in ElevenLabs' texts.	Bears on 21.6 Casting a voice, and who may speak in a final; 23.9 Limits, prices and rights (ElevenLabs). Listed in 21.11, 23.10 (ElevenLabs), 34.12.
Whether Eleven v4, or v3, is a beta service on your plan. If it is, its output cannot ship. The models and help pages carry no such label, but one app FAQ heading still reads "Eleven v3 (Alpha)".	Read the label before client use.	The label on your plan, or ElevenLabs in writing.	Bears on 23.9 Limits, prices and rights (ElevenLabs). Listed in 23.10 (ElevenLabs), 34.12.
Whether IPA works for Arabic, and whether a dictionary's phoneme rule applies on v4 or v3. Neither is documented.	Audition it on the failing word; use an alias for a recurring brand.	The audition on the failing word, or the maker's statement.	Bears on 23.3 The prompt, in this tool's dialect (ElevenLabs); 21.7 The Egyptian line workflow: repair, measure, lock. Listed in 23.10 (ElevenLabs).
What Voice Changer defaults to, and whether it reads Egyptian. The default model is not stated and Egyptian is not listed.	Name the model and hear a test.	A named model and a test heard by a native ear.	Bears on 23.2 How it reads what you give it (ElevenLabs); 21.3 Three roads to a voice. Listed in 23.10 (ElevenLabs).
What the cross-language accent rule does to a non-Arabic voice reading an Egyptian line. Not stated.	Cast a native voice.	A test of a non-Arabic voice on an Egyptian line, heard by a native ear.	Bears on 21.3 Three roads to a voice. Listed in 23.10 (ElevenLabs).
What a sound effect costs on ElevenLabs, and how short it can be. Three prices and two minimum durations are published.	Read the price before the job runs, and probe the file for its length.	The platform's quote for the job, and the first file.	Bears on 23.9 Limits, prices and rights (ElevenLabs). Listed in 23.10 (ElevenLabs).
Which plates and shapes pass on Sync Lipsync 3. A take head of a second or more, room tone in the pad, a trimmed plate joined to its own business frames, a sound-off Kling plate, a pro plate, a plate over 5 s, a take over about 4.7 s, silence, a 1080p plate, an upscale after the pass, a profile, a beard and a hand near the mouth have not been run.	Test on a scratch clip before a client shot depends on any of them.	One scratch clip of each.	Bears on 22.3 Design the sync out first; 24.3 The prompt, in this tool's dialect (Sync Lipsync 3). Listed in 16.10 (Kling), 22.9, 24.10 (Sync Lipsync 3).
A talking plate against a closed-mouth plate on Sync Lipsync 3. The maker's wording favours the talking plate.	Make a talking plate for the take (22.3).	One scratch pair on your own shot, judged on the returned mouths.	Bears on 22.4 The keyframe and the talking plate. Listed in 22.9.
Which sync-3 build Higgsfield runs, and whether it has the maker's wider face window. It cannot be settled from outside.	Watch the face around the mouth on every pass.	Higgsfield's answer.	Bears on 24.7 Checks before you sign (Sync Lipsync 3). Listed in 24.10 (Sync Lipsync 3).
Sync-3 on Egyptian and Saudi mouth shapes. The maker says nothing about dialects.	Judge every pass with a native eye and ear.	A native eye and ear on returned clips.	Bears on 24.2 How it reads what you give it (Sync Lipsync 3). Listed in 24.10 (Sync Lipsync 3).

G.7 Music, sound and loudness

Chapters 8, 25 to 28 and 30.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
Loudness for Egyptian, Saudi and Gulf broadcasters, cinema advertising and the social platforms. No standard covers them and no publisher's figure has been read.	Ask the delivery owner in writing, and record the answer (8.2, 30.6).	The delivery owner's written figure for each destination.	Bears on 8.2 Decisions made at the start of every project; 30.6 Sound in the finish, and loudness. Listed in 8.10, 25.17, 30.10.
Whether a video model given a cue's percussion lands its action on the accents. No maker documents it.	Treat any such clip as a rough guide, and place every hit in the edit.	A clip run on a cue with clear accents, and the landing checked frame by frame.	Bears on 25.4 Music-first or picture-first. Listed in 25.17.
Whether Suno holds a named figure across a cue. No maker states it.	Listen for it on every cue (26.7), and do not promise it to a client.	A maker's statement, or the same figure listened for across several cues.	Bears on 26.7 Checks before you sign (Suno). Listed in 26.10 (Suno).
Whether Voices, Personas and Studio run on Suno v6. The help pages do not say.	Read the model selector before you plan on one.	The model selector on your account.	Bears on 26.6 Settings (Suno). Listed in 26.10 (Suno).
What Cover, Add Vocals and Replace Section keep on v6, and whether stems are separated or regenerated. The pages are older than the model.	Run each once on a cue you can afford to lose and compare it with its parent.	That run of each.	Bears on 26.6 Settings (Suno). Listed in 26.10 (Suno).
Whether a music generator sings Egyptian or Saudi dialect correctly, Suno or Eleven Music. Not established.	Sung Arabic goes to a dialect reviewer word by word; a line that must be right goes to a singer.	A reviewer's judgement of sung lines from each tool.	Bears on 26.1 What it is (Suno); 26.5 Examples at the standard (Suno). Listed in 26.10 (Suno), 27.10 (Eleven Music).
The real field lengths in Suno, and Studio's export format. Only a third-party reading exists for the first.	Read the export format from the first file you export.	Suno's published limits, and that first export.	Bears on 26.6 Settings (Suno). Listed in 26.10 (Suno).
Whether the terms of 26 May 2026 cover Eleven Music 2.5, and whether Eleven Music 2.5 may be used for a commissioned film.	Ask in writing before any use beyond online.	ElevenLabs' written answer.	Bears on 27.9 Limits, prices and rights (Eleven Music). Listed in 27.10 (Eleven Music), 34.12.
How long an Eleven Music track may be. The overview says 5 minutes, the API 10.	Test before planning a long cue.	One long request, or the maker's clarification.	Bears on 27.9 Limits, prices and rights (Eleven Music). Listed in 27.10 (Eleven Music).
Firefly's contract terms for audio. The FAQ allows commercial use; the terms page for generative AI could not be read.	Read Adobe's terms before a client final.	Adobe's terms page, read by a person.	Bears on 28.9 Limits, prices and rights (sound-effect generators). Listed in 28.10 (sound-effect generators).
Whose revenue counts in the Stable Audio Community License. The licence page says "you or your organization" and, elsewhere, "your organization, or the organization(s) performing the commercial use".	If your production company, your agency or your client earns over USD 1M a year, ask Stability, or a lawyer, before a client final.	Stability's answer, or a lawyer's.	Bears on 28.9 Limits, prices and rights (sound-effect generators). Listed in 28.10 (sound-effect generators).
Lyria and Arabic. Arabic is not on the eight-language list read.	Do not brief an Arabic vocal there. Egyptian and Saudi speech comes from a voice actor or ElevenLabs (Chapter 21).	Google's published language list showing Arabic.	Bears on 28.1 What it is (sound-effect generators); 28.9 Limits, prices and rights (sound-effect generators). Listed in 28.10 (sound-effect generators).
The sample rate of a Clip file. Two pages give 44.1 and 48 kHz.	Probe the file.	The probe, or the maker's correction.	Bears on 28.6 Settings (sound-effect generators). Listed in 28.10 (sound-effect generators).

G.8 Edit, finish and handover

Chapters 29 to 32.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
Whether a J or L cut survives the interchange file. Overlaps through the XML are not shown to carry.	Build overlaps in Premiere, and test before relying on the file.	One imported passage of overlapping dialogue and music.	Bears on 29.7 Alternates, and J and L cuts; 31.5 Building and verifying a handoff. Listed in 29.14, 31.5, 32.10 (Premiere Pro).
Whether five-second requests keep returning 121 frames at other durations and on other models.	Probe every file.	Probes across durations and models.	Bears on 14.15 Format, length, resolution, handles and probing; 29.4 Logging a take by its usable interval. Listed in 29.14.
Whether to upscale before or after colour for generated footage. No maker states an order.	Upscale from neutral footage, and let a comparison on your own footage overrule.	A comparison on your own footage.	Bears on 30.3 Repair, then decide whether to upscale. Listed in 30.10.
Whether to upscale after lip sync, or to make the plate at the delivery size.	Test both on the hard interval.	That test.	Bears on 30.3 Repair, then decide whether to upscale. Listed in 30.10.
Safe-zone figures for any platform. None read from a publisher.	Ask for the overlay.	The platform's published overlay.	Bears on 30.8 Aspect ratios and versions. Listed in 30.10.
Whether FCPXML imports into Premiere 26.5. Adobe updated its page on 2 Apr 2026.	Test a small file before promising it.	That small file.	Bears on 31.5 Building and verifying a handoff. Listed in 31.5.
Whether Premiere's scripting interface, constant speed changes, a native Arabic text layer, OTIO, EDL and AAF survive a handoff. None is shown to.	Build natively, and test before relying.	A small handoff exercising each.	Bears on 31.5 Building and verifying a handoff. Listed in 31.5.
Whether the JSON transcript import serves Arabic captions. Its word-level schema is a separate file that has not been read.	Set the captions by the canonical line, not the transcript.	One Arabic transcript imported and timed against the waveform.	Bears on 30.7 Captions, and Arabic text. Listed in 32.10 (Premiere Pro).
The Text Engine setting. The path comes from Adobe community replies of 2021.	Confirm it on your build.	The setting found on your build.	Bears on 30.7 Captions, and Arabic text. Listed in 32.10 (Premiere Pro).
Whether the platform exposes the FLUX 3 Video Upscaler.	Check the connector's upscale options before planning around it.	That check.	Bears on 30.3 Repair, then decide whether to upscale. Listed in 32.10 (Premiere Pro).
Generative Extend's frame rate and range on HDR and 60 fps clips.	Ask for the length when you generate, not in the edit (29.11).	One test extension on an HDR clip and a 60 fps clip.	Bears on 32.9 Limits, prices and rights (Premiere Pro); 29.11 Back to generation. Listed in 32.10 (Premiere Pro).

G.9 Platform, money and rights

Chapters 33 and 34.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
Which ElevenLabs version runs under Higgsfield's speech engine.	Higgsfield's speech tools are for English scratch lines. The approved voice is made elsewhere and reaches the platform as an upload (33.9).	Higgsfield's answer.	Bears on 33.2 How it reads what you give it (Higgsfield). Listed in 33.10 (Higgsfield).
Whether the connector exposes Seedance Draft Mode (a 480p preview, in the changelog of 22 Sep 2026), whether two assistants share one workspace, and how many jobs it takes at once.	Send only what you can watch.	One listing on each.	Bears on 33.9 Limits, prices and rights (Higgsfield). Listed in 33.10 (Higgsfield).
Whether the model sees resized uploads. The job history stored uploaded references as resized JPEGs (27 Sep).	Run one fine-lettering reference as an upload and as a job number, and compare.	That comparison.	Bears on 33.2 How it reads what you give it (Higgsfield). Listed in 33.10 (Higgsfield).
Whether the developer API counts as "on the platform" for a production that needs Kling multi-shot.	The book teaches one route, an assistant through the connector, and treats Kling's multi-shot storyboards as one clip per job.	The author's decision.	Bears on 33.2 How it reads what you give it (Higgsfield).
Seedance at 1080p: 45 or 60 credits, and why.	Quote the job on the day.	Two quotes on the day, with and without the start frame.	Bears on 34.2 Checking the cost on the day. Listed in 34.12.
Prices not yet quoted: Seedance edit, extension, video input and audio-only; Kling Omni Edit; any 15 s shape; the video upscalers.	The first ledger line is the price, inside a packet with a cap.	The quote for each, on the day.	Bears on 34.2 Checking the cost on the day; 8.10 The spending packet and the budget. Listed in 34.12.
For a lawyer: copyright in generated work; portrait rights and consent wording; a dead public figure's name in a reconstruction; whether a commissioned advertisement is an evidently creative deepfake, and who must disclose it.	Ask counsel before a final rests on any of them.	Counsel's written advice.	Bears on 34.7 Real people, and the references you attach. Listed in 34.12.

G.10 The worked productions

Chapters 35 and 36.

Open	What the book teaches meanwhile	What would settle it	Bears on; listed in
The speaking-shot contract for the outcry. The trailer's table does not record whether the exact take shipped.	Your project states its contract at the start (8.2). Example 35.3 shows the exact-take route only.	A record of what shipped on the outcry.	Bears on 8.2 Decisions made at the start of every project; 22.1 What this stage decides, and who decides it. Listed in 35.11.
Whether the corrected prompts improve what comes back. They have not been run; a landing sentence, a measured second system or a wordless plate written this way is unproven.	Test one shot before you build a sequence on it.	That shot.	Bears on 35.11 What travels, and what stays with Mastorna. Listed in 35.11.
The train. It is the film's hardest clip.	Keep budget for variants, and expect to pin its end with a second frame.	The clip, made.	Bears on 14.8 Start and end frames. Listed in 35.11.
Whether a printed Latin label can be copied from a photograph of the pack.	Judge the round 3 jar frames against the real pack, glyph by glyph, once the six photographs exist.	The six photographs of the pack, and that comparison.	Bears on 10.11 Products: the real thing is the authority; 10.12 Text in the frame: two layers. Listed in 36.8.
Whether the pack was cleared for upload.	Until it is public and the client has agreed in writing, use a stand-in.	The client's written agreement.	Bears on 10.11 Products: the real thing is the authority. Listed in 36.8.
The cream's colour. Beige, as the client said on 22 Sep, or near-white, as the round 3 prompts said.	Ask.	The client's answer.	Bears on 10.11 Products: the real thing is the authority. Listed in 36.8.
The science claims. They are the brand's to support before any film states them.	Do not state them until the brand supports them.	The brand's substantiation.	Bears on 36.1 The brief and the decisions made at the start. Listed in 36.8.