File: Beetles (Great Southeast Pollinator Course, Week 7 of 7)
Date Corrected: 2026-07-31 (supersedes the 2026-07-30 pass)
Audio-Verification Pass: 2026-07-31, same session. Two of three flags resolved.
Webinar Date: July 28, 2026
Series: Great Southeast Pollinator Course
Topic: Entomology, beetle systematics and beetle pollination
Presenter: Dr. Joe McHugh, Ross Memorial Distinguished Professor of Insect Systematics and Evolutionary Biology, Department of Entomology, University of Georgia; Curator of Arthropods, Georgia Museum of Natural History
Host: Becky Griffin, National Coordinator, Great Southeast Pollinator Census
Co-host: Rich Braman, UGA Center for Urban Agriculture
Also speaking: Amanda Bratcher, NC Cooperative Extension, North Carolina State Coordinator
SOURCE VERIFICATION
This pass replaces the 2026-07-30 Stage 1 output. Rich reworked the audio and had it retranscribed twice, once through FireCut and once through MacWhisper. Both files were read in full before any editing.
| FireCut | MacWhisper | |
|---|---|---|
| Blocks | 989 | 296 |
| Lines | 3,956 | 1,184 |
| Time range | 00:00:00,100 to 01:13:54,769 | 00:00:06,239 to 01:13:49,300 |
| Speaker labels | none | yes, 4 speakers |
| Gaps over 5 seconds | none | three (16:27, 30:25, 32:34) |
| Overlapping blocks | none | seven |
| Spoken words | 11,886 | 11,317 |
- Backbone selected: FireCut, on Rich’s instruction and confirmed by measurement.
- Corrected blocks: 989 original = 989 corrected ✓ MATCH CONFIRMED
- Corrected lines: 3,956 original = 3,956 corrected ✓
- Time range: 00:00:00,100 to 01:13:54,769 (~73.9 min)
- File reading: COMPLETE ✓ (both sources, full read, no sampling)
- Speaker distribution after labeling: McHugh 631 blocks, Griffin 293, Braman 43, Bratcher 21, one unlabeled block (the closing music cue).
Coverage proof:
- Early [00:03:55, Block 67]: McHugh names his own specialism, “I am a systematist, and systematists are people who study the science of biodiversity.”
- Middle [00:33:39, Block 462]: “So mess and soil pollinators,” closing the cantharophily section on beetles as dust mop and tractor at once.
- Late [01:10:12, Block 944]: Amanda Bratcher reports that North Carolina has been doubling its census participation since joining in 2023.
Why Two Sources, and What Each Contributed
The two engines fail in different places, so the pass used each for what it is good at.
FireCut supplied the text. It agrees with MacWhisper on 94.5 percent of words but carries about 570 more of them: MacWhisper silently drops five stretches of speech, the longest sixteen words. FireCut also produced no gaps, no overlapping timestamps, and 989 short blocks against MacWhisper’s 296 paragraph blocks, which gives Stage 2 chapters and Stage 4 timestamp references far finer granularity than any previous week of this course. Most important for a taxonomy session, FireCut resolved on its own almost every family name that the original Whisper pass had to have corrected by hand: Canidae, Chordata, Plantae, Carnivora, Staphylinidae, Lampyridae, Scarabaeidae, Cerambycidae, Chrysomelidae, Cantharidae, Curculionidae, Mordellidae, plus Systema Naturae, Linnaean, Müllerian, cantharophily, cantharidin, posteromotorism, setae, holometabolous, pronotum, tarsi, elytra, dorsoventrally, and the full nine-item list of feeding categories from frugivorous to detritivorous. It also got Vergennes, Vermont, Duane McKenna, bess beetles, maxillae, and the Arthur Evans book title right, all of which were errors or open flags last time.
MacWhisper supplied the speaker labels, and it resolved one term FireCut lost. Where FireCut heard “Julidae” for the sap beetle family, MacWhisper heard “nitidulidae,” which settles the family with no outside knowledge required. That single cross-check accounts for seven of this pass’s corrections.
Neither engine got the click beetle family. FireCut heard “Piridae,” MacWhisper heard “Carity.” A third engine was run against the audio during the verification pass and did not get it either. That flag stays open.
The block 209 data loss from the previous pass is gone. The old Whisper file had an ASR loop repeating “Community” 58 times and lost 17.3 seconds of Griffin’s Facebook-page remarks. Both new transcripts cover that stretch continuously, and the content is intact in the corrected file at blocks 789 through 805. No hand transcription is needed and no rendering convention has to be invented.
Corrections Applied
Twenty-three corrections across 22 of 989 blocks, after the audio-verification pass closed two flags without needing a text change. That is roughly one correction per three minutes of runtime, against 84 corrections across 57 blocks in the 2026-07-30 pass. The drop is the point: the reworked audio and the better engine removed most of the work rather than moving it.
Scientific Names
The sap beetle family, resolved by cross-source agreement. FireCut renders it “Julidae” once and then “nidulid” in six forms; MacWhisper independently renders it “nitidulidae.” McHugh names the common names (“sap beetles, picnic beetles, and pollen beetles”) and the diagnostic characters (lollipop-on-a-stick antennae, truncate elytra, pawpaw pollination) in the same passage, so the identification is settled inside the transcript.
- “Julidae are the sap beetles” → “Nitidulidae are the sap beetles” (Block 590)
- “for the nidulids” → “for the nitidulids” (Block 598)
- “The elytra nidulids are truncated.” → “The elytra in nitidulids are truncated.” (Block 600, MacWhisper also supplies the dropped preposition)
- “So nidulidae, very common.” → “So Nitidulidae, very common.” (Block 602)
- “By the way, nidulids are a family.” → “By the way, nitidulids are a family.” (Block 603)
- “They’re, they’re nidulid beetles” → “They’re, they’re nitidulid beetles” (Block 604)
- “it’s nidulids that are” → “it’s nitidulids that are” (Block 606)
Technical Terms
- “usually the four wings are the bigger, more powerful wings” → “usually the forewings are the bigger, more powerful wings” (Block 128). Both engines produced the homophone. Block 127 has already established the count as four, which makes “the four wings are the bigger wings” say nothing; Block 137 restates the same idea as “the anterior pair of wings, which are usually the major wings that power the flight”; and Block 157 uses “forewings” outright. This one matters, because it sits in the elytra passage that Stages 2, 4 and 5 will all draw on.
- “posteriomotorism” → “posteromotorism” (Block 147). The standard Coleoptera term for flight powered by the hind wings alone. MacWhisper heard “posterior motorism.”
- “This is a Lepidoptera” → “This is a lepidopteran” (Block 354). Established course pattern from Week 6; MacWhisper heard “lepidopter.”
- “In the beetle verse” → “In the beetleverse” (Block 368) and “in the beetle verse” → “in the beetleverse” (Block 392). Standardized to the speaker’s own coinage as FireCut renders it at Blocks 331, 333 and 339.
- “those mandibulates mouthparts” → “those mandibulate mouthparts” (Block 453). Matches Blocks 178 and 454.
- “and they can’t bioluminescence” → “and they can’t bioluminesce” (Block 561). Block 562 repeats the clause correctly, and MacWhisper has the verb in both positions.
Grammar
- “But this, of course, has just been one for our history books, Becky.” → “But this course has just been one for our history books, Becky.” (Block 6). MacWhisper heard “this course,” which is what Braman is talking about: this is the final session of the Great Southeast Pollinator Course. FireCut’s comma-bracketed “of course” leaves the sentence without a subject noun for “one for our history books.”
- “let’s talk about the biology” → “Let’s talk about the biology” (Block 412). Sentence-initial capitalization.
- “we have one of the flower beetles” → “We have one of the flower beetles” (Block 700). Sentence-initial capitalization.
- “like at twelve O one AM” → “like at twelve oh one a.m.” (Block 773). Time-of-day rendering.
Plant Names
- “on a jewel weed” → “on a jewelweed” (Block 700). Standardized compound, consistent with the course convention for compound common names.
Proper Nouns
- “at the governor’s mansion again this year” → “at the Governor’s Mansion again this year” (Block 922). Proper noun. “Governor’s wife” (Block 923) and “the governor” (Block 925) stay lowercase, correctly.
Institutions and Programs
- “your local extension personnel” → “your local Extension personnel” (Block 930). Cooperative Extension is capitalized.
Transcription Artifact
- “(outro music)” → “(intro music)” (Block 1). FireCut tags the 0:00 pre-roll as outro music. It is the Beatles warm-up music Griffin refers to a minute later, and the genuine outro tag is at Block 989. The cue was also moved outside the speaker label, so the file does not attribute a sound effect to Rich Braman.
FireCut’s other non-speech cues, “(laughs),” “(coughs),” “(chuckles),” were kept as transcribed. This is new for the course, since the Zoom Whisper exports never produced them. They are source content, and Stage 5 can use them or drop them.
Resolved on Audio-Verification Pass (2026-07-31)
Two clips were cut from the Resolve render, Great Southeast Pollinator Course - Week Seven - Beetles.mov (4,434.90 s, matching the FireCut source to the millisecond), and put through a third engine, Whisper large-v3, under several priming conditions. Two flags closed. Neither needed a text change, which is the useful outcome: the file already said what the speaker said.
Block 21, “Jean A.” RESOLVED, no change. This one could never have been settled by ear, because Jean and Gene are homophones. It is a spelling question, so it was resolved documentarily instead: UGA’s own announcement names the gift the “Herbert H. and Jean A. Ross Memorial Endowment,” and the Entomology department lists McHugh’s chair as the “Herbert H. and Jean A. Ross Memorial Distinguished Professor of Insect Systematics and Evolutionary Biology.” FireCut had it right; MacWhisper’s “Gene A.” was the error. Worth recording as a general point: a homophone flag is not an audio flag, and sending it to the audio pass wastes a listen.
Block 21, “System- System.” RESOLVED, no change. Griffin does not say “Systematics.” Whisper large-v3 transcribes the phrase as “Insect Systems and Evolutionary Biology” with an average log probability of -0.053 and word probabilities of 1.00 on every neighbouring word, which is as clean as this engine gets. That is three independent engines agreeing on “System” or “Systems,” and none hearing a “-atics” ending. FireCut’s “System- System” additionally captures the stumble, which the other two engines smooth away. The spoken form stands. The official chair title is “Insect Systematics and Evolutionary Biology,” and that is the form derived deliverables should use in a byline; the transcript records that Griffin misspoke it, which is what a transcript is for.
Flagged for Verification: 1 Open
The flag keeps the original ASR token inside the marker, per the course convention set on this webinar, and asserts no expected answer inside the SRT, so no unspoken word enters the source document.
Block 579, “Piridae” at 00:43:29,560. Piridae [VERIFY: "Piridae"; click beetle family name] are the click beetles. Still the only one of the eight families in the identification tour whose name did not survive transcription, and Stage 2 chapter titles, the Stage 4 matching exercise and any Stage 5 section heading will want it.
Three engines, three different answers, and the third one will not be moved. FireCut heard “Piridae,” MacWhisper heard “Carity,” Whisper large-v3 heard “Charity” at 0.68 probability. The interesting result is what happened under priming. Given an initial prompt containing ten beetle family names including Elateridae, large-v3 still returned “Charity.” Given the literal sentence “Elateridae are the click beetles” as its prompt, it still returned “Charity are the click beetles.” A wrong-target control prompt produced exactly the same output, so the model is not being pulled by the prompt in either direction. Every surrounding word in that segment scores 1.00, so the audio itself is clean; it is this one token that no engine can place.
That is a strong negative result rather than a failure to try. It means the machine path is exhausted and the acoustics do not obviously support Elateridae, so writing Elateridae in would be a guess dressed as a correction. This one needs a human ear. The clip is at gtbop-staging/_audio_verify_week7/flagB_block579_4322-4338.wav; the word lands about 7.4 seconds in, immediately before “are the click beetles.”
Not Corrected, but Worth Knowing
None of the following is an ASR error, so none was changed. All of them can mislead a downstream stage.
- Block 300, “ventral nerve cord.” McHugh says chordates have a ventral nerve cord. Chordates are defined by a dorsal nerve cord; ventral is the arthropod condition. Both engines agree on the word, so this is a speaker slip, not a transcription error, and it stays as spoken. It must not be repeated as fact in Stage 2, Stage 4 or Stage 5.
- Blocks 296 and 302, duplicated species figure. Class Mammalia is given 5,000 species and phylum Chordata is given the same 5,000. Chordata contains Mammalia, and the escalating comparison (35, then 270, then 5,000, against 65,000 weevils) needs the phylum figure to be larger. Both engines agree, so this is speech or a slide, not ASR. Check the slide before writing a quiz question on either number. Block 309’s stumbled figure for kingdom Plantae, “thirty-two, uh, two hundred, uh, three hundred and twenty thousand,” resolves to 320,000.
- Block 362, “thorax” against “prothorax.” FireCut has “these aren’t horns on the head or horns on the thorax,” MacWhisper has “prothorax.” Beetle horns arise on the head and the pronotum, and McHugh uses “pronotum” precisely six times elsewhere, so “prothorax” may be the spoken word. FireCut is kept as the backbone. Worth a listen if Stage 4 builds a horn morphology item.
- Block 827, “five hund- ninety-four schools.” The stutter is preserved. The figure is 594.
- Block 854, “dedicated to the Walter Reeves.” Both engines have the stray article. Left as spoken.
- Block 626, the book title. McHugh says “this book here by Arthur Evans called The Beetles of Eastern North America.” The published title is Beetles of Eastern North America by Arthur V. Evans. Cite the published form in the Stage 2 resources block, not the spoken one.
- Blocks 244 and 245, the 150 million year figure. FireCut attaches “over a hundred and fifty million years ago” to the cycad forests; MacWhisper attaches it to the following clause about pre-bee pollination. The facts are the same either way, but a chapter title or a recall question should not lean on the attachment.
- Names appearing once, spelling unverified by a second source: Megan McCoy (Block 736, credited with all the census mascot artwork), Dr. Levy (Block 919, surname only, no affiliation given), and Jeremy (Block 943, first name only, apparently the Florida state coordinator). Both engines agree on each, so none is flagged, but each will be quoted in a public deliverable.
Speaker Attribution
MacWhisper’s diarization produced only fifteen turns across 74 minutes, and it is demonstrably wrong in several places: it puts McHugh’s “I’m a beetle guy, I’m not a plant guy” in Griffin’s voice, and it folds McHugh’s opening screen-share fumbling into Griffin’s introduction. Labels were therefore assigned by content and cross-checked against the diarization rather than copied from it.
One cross-check turned out to be reliable and is worth recording for future weeks. MacWhisper’s single-stream ASR drops overlapped backchannels entirely, while it transcribes a speaker’s own discourse markers inline. Every short interjection that content analysis identified as cross-talk is missing from MacWhisper: “Right.” (Block 891), “Okay.” (896), “Okay.” (902), “Great.” (947), “That’s awesome.” (962). That omission test separates a genuine interruption from a speaker clearing his own throat, and it caught five turn boundaries that a straight timestamp merge would have buried inside someone else’s run.
Mixed-Speaker Blocks: 3
The speaker changes inside these blocks, so no single label is fully correct. All three are assigned to the speaker who holds most of the block. They need attribution by content at Stage 5 and cannot be fixed in the SRT without splitting blocks, which the protocol forbids.
- Block 661 (Griffin): Griffin reads Wendy’s coontie question and the block ends on “I’m,” which is the first word of McHugh’s reply in Block 662.
- Block 669 (McHugh): “I’m not a, I’m not a plant- Not a plant guy Sorry.” Griffin finishes his sentence for him inside the block.
- Block 895 (Braman): “Maybe it got forked a little bit in the- Oh, got tw- In the slide.” Braman and Griffin talk over each other about the mascot QR code. “Forked” is what both engines heard, so it is left as transcribed rather than flagged.
Open Speaker Questions for the Audio Pass
Three low-stakes attributions that a listen would settle. None changes any claim in the transcript.
- Block 608, “Mm.” (00:45:42.5), currently McHugh. It is 80 milliseconds long, sits in a clean 200 millisecond gap between two of his completed sentences, and MacWhisper drops it entirely, which is the cross-talk signature described above. McHugh also uses “Okay?” and “All right?” as his discourse tags throughout and never uses “Mm.” elsewhere. It may be Griffin acknowledging the new family.
- Block 35, “You might need to sh…” (00:02:07), currently Braman. MacWhisper’s diarization puts this fragment on Griffin’s side of a speaker boundary. It is assigned to Braman because he handles the screen-share exchange either side of it (Blocks 36 to 38, then 45 to 48), which makes it one continuous tech-support turn.
- Block 893, “Okay.” (01:07:00), currently Braman. Reassigned from Griffin during audit: MacWhisper retains it inline in Braman’s stream while dropping the confirmed Griffin backchannels around it, and “Okay, we’re saying maybe the QR code” reads as one thought.
SRT Format Compliance
✅ All 989 timestamps preserved exactly as in the FireCut source, verified by full-file diff, not sampling
✅ All 989 sequence numbers unchanged
✅ Blank lines between segments preserved; line count 3,956 in, 3,956 out
✅ No segments merged, split, added or dropped
✅ Block count: 989 original = 989 corrected ✓
✅ Every block carries exactly one text line, none empty
✅ Only the 23 declared blocks differ from the source text, plus the speaker label prefix
Method was the course standard: exact string replacement against the raw file with a uniqueness assertion on every pattern, then a full-file diff of all sequence and timestamp lines. Structural drift is zero.
Line length. FireCut’s blocks are single lines of up to about 130 characters, which exceeds the GTBOP guideline of two lines at roughly 42 characters. This is inherited from the source and is the same shape every previous week of this course has used. Reflowing would require re-splitting the text, which the protocol forbids at Stage 1. Noted rather than fixed.
New Correction Patterns for the Course README
FireCut and MacWhisper fail differently from Zoom Whisper, so most of the 41 patterns added on 2026-07-30 did not recur. These are the new ones.
| Engine output | Correct form |
|---|---|
| Julidae / nidulids / nidulidae (FireCut) | Nitidulidae / nitidulids |
| Piridae (FireCut) / Carity (MacWhisper) / Charity (Whisper large-v3) | click beetle family, still flagged after three engines |
| Gene A. Ross (MacWhisper) | Jean A. Ross (UGA endowment name) |
| four wings (where forewings is meant) | forewings |
| posteriomotorism / posterior motorism | posteromotorism |
| mandibulates mouthparts | mandibulate mouthparts |
| can’t bioluminescence | can’t bioluminesce |
| Lepidoptera (naming a single moth) | lepidopteran |
| beetle verse | beetleverse |
| jewel weed | jewelweed |
| this, of course (where “this course” is meant) | this course |
| twelve O one AM | twelve oh one a.m. |
| (outro music) tagged at 00:00:00 | (intro music) |
Also worth recording as a method note: on this recording FireCut resolved unaided almost every taxon name that Zoom Whisper mangled, including Canidae, Chordata, Plantae, Carnivora, Staphylinidae, Lampyridae, Cerambycidae, Chrysomelidae, Cantharidae, Curculionidae, Systema Naturae, Müllerian, cantharophily and cantharidin. If a future taxonomy-heavy session is available in both engines, run FireCut for the text.
Total Corrections: 23 individual corrections across 22 blocks
Flags: 3 raised, 2 resolved on the audio-verification pass, 1 open (Block 579, the click beetle family)
Processing: Complete file, 989 subtitle blocks, 3,956 lines, both sources read in full
Audio verification used the Resolve render at Movies/Davinci/Renders/Great Southeast Pollinator Course/Great Southeast Pollinator Course - Week Seven - Beetles.mov. Clips retained at gtbop-staging/_audio_verify_week7/.
Sources for the endowed-chair name: UGA Today, endowment honoring Herbert H. and Jean A. Ross and UGA Department of Entomology faculty page for Joseph V. McHugh.
Correction summary produced for UGA Center for Urban Agriculture / Great Southeast Pollinator Course Archives
Source: FireCut SRT, “Firecut GreatSoutheastPollinatorCourseWeekSevenBeetles.mov.srt” (989 blocks), speaker labels merged from MacWhisper SRT, “Macwhisper Great Southeast Pollinator Course Week Seven Beetles.srt” (296 blocks)
Supersedes the 2026-07-30 correction pass built on the original Zoom Whisper export (259 blocks)