diff --git a/surfsense_backend/app/podcasts/api/routes.py b/surfsense_backend/app/podcasts/api/routes.py
index f55febcbd..f2d0c6b47 100644
--- a/surfsense_backend/app/podcasts/api/routes.py
+++ b/surfsense_backend/app/podcasts/api/routes.py
@@ -157,8 +157,8 @@ async def create_podcast(
         session,
         search_space_id=body.search_space_id,
         speaker_count=body.speaker_count,
-        min_minutes=body.min_minutes,
-        max_minutes=body.max_minutes,
+        min_seconds=body.min_seconds,
+        max_seconds=body.max_seconds,
         focus=body.focus,
     )
     await service.attach_brief(podcast, spec)
diff --git a/surfsense_backend/app/podcasts/api/schemas.py b/surfsense_backend/app/podcasts/api/schemas.py
index 7f1f8cc7c..72c2a5f7a 100644
--- a/surfsense_backend/app/podcasts/api/schemas.py
+++ b/surfsense_backend/app/podcasts/api/schemas.py
@@ -11,6 +11,12 @@ from datetime import datetime
 
 from pydantic import BaseModel, ConfigDict, Field
 
+from app.podcasts.duration_limits import (
+    DEFAULT_MAX_SECONDS,
+    DEFAULT_MIN_SECONDS,
+    MAX_DURATION_SECONDS,
+    MIN_DURATION_SECONDS,
+)
 from app.podcasts.persistence import Podcast, PodcastStatus
 from app.podcasts.schemas import PodcastSpec, Transcript
 from app.podcasts.service import has_stored_episode, read_spec, read_transcript
@@ -18,8 +24,6 @@ from app.podcasts.service import has_stored_episode, read_spec, read_transcript
 # Defaults applied when a create request omits brief sizing; the brief gate lets
 # the user adjust before any cost is incurred.
 DEFAULT_SPEAKER_COUNT = 2
-DEFAULT_MIN_MINUTES = 10
-DEFAULT_MAX_MINUTES = 20
 
 
 class CreatePodcastRequest(BaseModel):
@@ -30,8 +34,16 @@ class CreatePodcastRequest(BaseModel):
     source_content: str = Field(..., min_length=1)
     thread_id: int | None = None
     speaker_count: int = Field(default=DEFAULT_SPEAKER_COUNT, ge=1, le=6)
-    min_minutes: int = Field(default=DEFAULT_MIN_MINUTES, ge=1)
-    max_minutes: int = Field(default=DEFAULT_MAX_MINUTES, ge=1)
+    min_seconds: int = Field(
+        default=DEFAULT_MIN_SECONDS,
+        ge=MIN_DURATION_SECONDS,
+        le=MAX_DURATION_SECONDS,
+    )
+    max_seconds: int = Field(
+        default=DEFAULT_MAX_SECONDS,
+        ge=MIN_DURATION_SECONDS,
+        le=MAX_DURATION_SECONDS,
+    )
     focus: str | None = Field(default=None, max_length=2000)
 
 
diff --git a/surfsense_backend/app/podcasts/duration_limits.py b/surfsense_backend/app/podcasts/duration_limits.py
new file mode 100644
index 000000000..fc7d29890
--- /dev/null
+++ b/surfsense_backend/app/podcasts/duration_limits.py
@@ -0,0 +1,6 @@
+"""Shared bounds and defaults for podcast target duration."""
+
+MAX_DURATION_SECONDS = 24 * 60 * 60
+MIN_DURATION_SECONDS = 15
+DEFAULT_MIN_SECONDS = 20
+DEFAULT_MAX_SECONDS = 30
diff --git a/surfsense_backend/app/podcasts/generation/brief/config.py b/surfsense_backend/app/podcasts/generation/brief/config.py
index 4f92585ae..9b206bde4 100644
--- a/surfsense_backend/app/podcasts/generation/brief/config.py
+++ b/surfsense_backend/app/podcasts/generation/brief/config.py
@@ -6,10 +6,13 @@ from dataclasses import dataclass, field, fields
 
 from langchain_core.runnables import RunnableConfig
 
+from app.podcasts.duration_limits import (
+    DEFAULT_MAX_SECONDS,
+    DEFAULT_MIN_SECONDS,
+)
+
 # Sensible defaults for a fresh brief; the user adjusts the range at the gate.
 DEFAULT_SPEAKER_COUNT = 2
-DEFAULT_MIN_MINUTES = 10
-DEFAULT_MAX_MINUTES = 20
 
 
 @dataclass(kw_only=True)
@@ -17,8 +20,8 @@ class BriefConfig:
     """Signals used to propose a brief; everything here is non-LLM context."""
 
     speaker_count: int = DEFAULT_SPEAKER_COUNT
-    min_minutes: int = DEFAULT_MIN_MINUTES
-    max_minutes: int = DEFAULT_MAX_MINUTES
+    min_seconds: int = DEFAULT_MIN_SECONDS
+    max_seconds: int = DEFAULT_MAX_SECONDS
     focus: str | None = None
     last_used_language: str | None = None
     last_used_voices: list[str] = field(default_factory=list)
diff --git a/surfsense_backend/app/podcasts/generation/brief/nodes.py b/surfsense_backend/app/podcasts/generation/brief/nodes.py
index c0a6f1ae1..de6a9717e 100644
--- a/surfsense_backend/app/podcasts/generation/brief/nodes.py
+++ b/surfsense_backend/app/podcasts/generation/brief/nodes.py
@@ -79,7 +79,7 @@ def propose_spec(state: BriefState, config: RunnableConfig) -> dict[str, Any]:
         style=PodcastStyle.CONVERSATIONAL,
         speakers=speakers,
         duration=DurationTarget(
-            min_minutes=brief.min_minutes, max_minutes=brief.max_minutes
+            min_seconds=brief.min_seconds, max_seconds=brief.max_seconds
         ),
         focus=brief.focus,
     )
diff --git a/surfsense_backend/app/podcasts/generation/brief/propose.py b/surfsense_backend/app/podcasts/generation/brief/propose.py
index 17344702b..09d74840e 100644
--- a/surfsense_backend/app/podcasts/generation/brief/propose.py
+++ b/surfsense_backend/app/podcasts/generation/brief/propose.py
@@ -4,11 +4,12 @@ from __future__ import annotations
 
 from sqlalchemy.ext.asyncio import AsyncSession
 
+from app.podcasts.duration_limits import DEFAULT_MAX_SECONDS, DEFAULT_MIN_SECONDS
 from app.podcasts.persistence import PodcastRepository
 from app.podcasts.schemas import PodcastSpec
 from app.podcasts.service import preferences_from
 
-from .config import DEFAULT_MAX_MINUTES, DEFAULT_MIN_MINUTES, DEFAULT_SPEAKER_COUNT
+from .config import DEFAULT_SPEAKER_COUNT
 from .graph import graph as brief_graph
 from .state import BriefState
 
@@ -18,8 +19,8 @@ async def propose_brief(
     *,
     search_space_id: int,
     speaker_count: int = DEFAULT_SPEAKER_COUNT,
-    min_minutes: int = DEFAULT_MIN_MINUTES,
-    max_minutes: int = DEFAULT_MAX_MINUTES,
+    min_seconds: int = DEFAULT_MIN_SECONDS,
+    max_seconds: int = DEFAULT_MAX_SECONDS,
     focus: str | None = None,
 ) -> PodcastSpec:
     """Reuse the last-used language and voices, else English; return the spec."""
@@ -29,8 +30,8 @@ async def propose_brief(
     config = {
         "configurable": {
             "speaker_count": speaker_count,
-            "min_minutes": min_minutes,
-            "max_minutes": max_minutes,
+            "min_seconds": min_seconds,
+            "max_seconds": max_seconds,
             "focus": focus,
             "last_used_language": last_language,
             "last_used_voices": last_voices,
diff --git a/surfsense_backend/app/podcasts/generation/transcript/nodes.py b/surfsense_backend/app/podcasts/generation/transcript/nodes.py
index 44d6b219d..7b472348d 100644
--- a/surfsense_backend/app/podcasts/generation/transcript/nodes.py
+++ b/surfsense_backend/app/podcasts/generation/transcript/nodes.py
@@ -38,7 +38,7 @@ async def plan_outline(
     tc = TranscriptConfig.from_runnable_config(config)
     llm = await _require_llm(state, tc)
 
-    target_words = round(tc.spec.duration.midpoint_minutes * _WORDS_PER_MINUTE)
+    target_words = round(tc.spec.duration.midpoint_seconds * _WORDS_PER_MINUTE / 60)
     suggested_segments = max(1, round(target_words / _WORDS_PER_SEGMENT))
 
     messages = [
diff --git a/surfsense_backend/app/podcasts/schemas/spec.py b/surfsense_backend/app/podcasts/schemas/spec.py
index 1ef3dcfff..3799d883b 100644
--- a/surfsense_backend/app/podcasts/schemas/spec.py
+++ b/surfsense_backend/app/podcasts/schemas/spec.py
@@ -10,17 +10,19 @@ from __future__ import annotations
 
 import re
 from enum import StrEnum
+from typing import Any
 
 from pydantic import BaseModel, ConfigDict, Field, field_validator, model_validator
 
+from app.podcasts.duration_limits import (
+    MAX_DURATION_SECONDS,
+    MIN_DURATION_SECONDS,
+)
+
 # A speaker count beyond this is almost never a real podcast and explodes the
 # voice/turn-attribution space, so we reject it at the brief gate.
 MAX_SPEAKERS = 6
 
-# Long-form is a goal, but an open-ended upper bound invites runaway TTS bills.
-# One day of audio is a generous ceiling that still blocks obvious mistakes.
-MAX_DURATION_MINUTES = 24 * 60
-
 # BCP-47 primary subtag plus optional region (e.g. ``en``, ``en-US``, ``pt-BR``).
 # Kept deliberately permissive: the voice catalog, not the brief, decides which
 # languages can actually be synthesised. Casing is normalised after matching.
@@ -91,7 +93,7 @@ class SpeakerSpec(BaseModel):
 
 
 class DurationTarget(BaseModel):
-    """The desired finished length as an inclusive minute range.
+    """The desired finished length as an inclusive second range.
 
     Drafting aims for the midpoint and treats the bounds as soft guardrails;
     storing a range (rather than a point) keeps long-form expectations honest
@@ -100,19 +102,38 @@ class DurationTarget(BaseModel):
 
     model_config = ConfigDict(extra="forbid")
 
-    min_minutes: int = Field(..., ge=1, le=MAX_DURATION_MINUTES)
-    max_minutes: int = Field(..., ge=1, le=MAX_DURATION_MINUTES)
+    min_seconds: int = Field(..., ge=MIN_DURATION_SECONDS, le=MAX_DURATION_SECONDS)
+    max_seconds: int = Field(..., ge=MIN_DURATION_SECONDS, le=MAX_DURATION_SECONDS)
+
+    @model_validator(mode="before")
+    @classmethod
+    def _coerce_legacy_minutes(cls, data: Any) -> Any:
+        """Rows stored before seconds-based briefs still load from JSONB."""
+        if (
+            isinstance(data, dict)
+            and "min_seconds" not in data
+            and "min_minutes" in data
+        ):
+            migrated = dict(data)
+            migrated["min_seconds"] = int(migrated.pop("min_minutes")) * 60
+            migrated["max_seconds"] = int(migrated.pop("max_minutes")) * 60
+            return migrated
+        return data
 
     @model_validator(mode="after")
     def _check_order(self) -> DurationTarget:
-        if self.max_minutes < self.min_minutes:
-            raise ValueError("max_minutes must be >= min_minutes")
+        if self.max_seconds < self.min_seconds:
+            raise ValueError("max_seconds must be >= min_seconds")
         return self
 
     @property
-    def midpoint_minutes(self) -> float:
+    def midpoint_seconds(self) -> float:
         """The runtime drafting should aim for within the range."""
-        return (self.min_minutes + self.max_minutes) / 2
+        return (self.min_seconds + self.max_seconds) / 2
+
+    @property
+    def midpoint_minutes(self) -> float:
+        return self.midpoint_seconds / 60
 
 
 class PodcastSpec(BaseModel):
diff --git a/surfsense_backend/tests/integration/podcasts/conftest.py b/surfsense_backend/tests/integration/podcasts/conftest.py
index 240f57f96..75248a6a1 100644
--- a/surfsense_backend/tests/integration/podcasts/conftest.py
+++ b/surfsense_backend/tests/integration/podcasts/conftest.py
@@ -217,7 +217,7 @@ def build_spec(
                 slot=1, name="Guest", role=SpeakerRole.GUEST, voice_id=voice_ids[1]
             ),
         ],
-        duration=DurationTarget(min_minutes=10, max_minutes=20),
+        duration=DurationTarget(min_seconds=600, max_seconds=1200),
     )
 
 
diff --git a/surfsense_backend/tests/integration/podcasts/test_draft_task.py b/surfsense_backend/tests/integration/podcasts/test_draft_task.py
index 7dadfc2f5..014d98b1f 100644
--- a/surfsense_backend/tests/integration/podcasts/test_draft_task.py
+++ b/surfsense_backend/tests/integration/podcasts/test_draft_task.py
@@ -76,8 +76,7 @@ async def test_quota_denial_fails_the_podcast_without_a_transcript(
     async def _deny(**_kwargs):
         raise QuotaInsufficientError(
             usage_type="podcast_generation",
-            used_micros=5_000_000,
-            limit_micros=5_000_000,
+            balance_micros=5_000_000,
             remaining_micros=0,
         )
         yield  # pragma: no cover - unreachable, satisfies the CM protocol
diff --git a/surfsense_backend/tests/unit/podcasts/conftest.py b/surfsense_backend/tests/unit/podcasts/conftest.py
index 5eb4d8457..c77eb1cc6 100644
--- a/surfsense_backend/tests/unit/podcasts/conftest.py
+++ b/surfsense_backend/tests/unit/podcasts/conftest.py
@@ -31,8 +31,8 @@ def make_spec():
         language: str = "en",
         style: PodcastStyle = PodcastStyle.CONVERSATIONAL,
         speakers: list[SpeakerSpec] | None = None,
-        min_minutes: int = 10,
-        max_minutes: int = 20,
+        min_seconds: int = 600,
+        max_seconds: int = 1200,
         focus: str | None = None,
     ) -> PodcastSpec:
         if speakers is None:
@@ -54,7 +54,7 @@ def make_spec():
             language=language,
             style=style,
             speakers=speakers,
-            duration=DurationTarget(min_minutes=min_minutes, max_minutes=max_minutes),
+            duration=DurationTarget(min_seconds=min_seconds, max_seconds=max_seconds),
             focus=focus,
         )
 
diff --git a/surfsense_backend/tests/unit/podcasts/test_renderer.py b/surfsense_backend/tests/unit/podcasts/test_renderer.py
index 2bcdff967..bb7b8f181 100644
--- a/surfsense_backend/tests/unit/podcasts/test_renderer.py
+++ b/surfsense_backend/tests/unit/podcasts/test_renderer.py
@@ -66,7 +66,7 @@ def _spec(voice_id: str) -> PodcastSpec:
         speakers=[
             SpeakerSpec(slot=0, name="Host", role=SpeakerRole.HOST, voice_id=voice_id)
         ],
-        duration=DurationTarget(min_minutes=5, max_minutes=10),
+        duration=DurationTarget(min_seconds=300, max_seconds=600),
     )
 
 
diff --git a/surfsense_backend/tests/unit/podcasts/test_spec.py b/surfsense_backend/tests/unit/podcasts/test_spec.py
index 4efd530e9..77e720286 100644
--- a/surfsense_backend/tests/unit/podcasts/test_spec.py
+++ b/surfsense_backend/tests/unit/podcasts/test_spec.py
@@ -57,7 +57,7 @@ def test_spec_normalizes_its_language_on_construction():
     spec = PodcastSpec(
         language="EN-us",
         speakers=[_speaker(0)],
-        duration=DurationTarget(min_minutes=5, max_minutes=10),
+        duration=DurationTarget(min_seconds=300, max_seconds=600),
     )
     assert spec.language == "en-us"
 
@@ -68,7 +68,7 @@ def test_speakers_must_have_unique_slots():
         PodcastSpec(
             language="en",
             speakers=[_speaker(0), _speaker(0, voice_id="kokoro:af_bella")],
-            duration=DurationTarget(min_minutes=5, max_minutes=10),
+            duration=DurationTarget(min_seconds=300, max_seconds=600),
         )
 
 
@@ -77,7 +77,7 @@ def test_a_brief_needs_at_least_one_speaker():
         PodcastSpec(
             language="en",
             speakers=[],
-            duration=DurationTarget(min_minutes=5, max_minutes=10),
+            duration=DurationTarget(min_seconds=300, max_seconds=600),
         )
 
 
@@ -86,7 +86,7 @@ def test_a_monologue_brief_carries_exactly_one_speaker():
         language="en",
         style=PodcastStyle.MONOLOGUE,
         speakers=[_speaker(0)],
-        duration=DurationTarget(min_minutes=5, max_minutes=10),
+        duration=DurationTarget(min_seconds=300, max_seconds=600),
     )
     assert spec.style is PodcastStyle.MONOLOGUE
 
@@ -98,18 +98,25 @@ def test_a_monologue_brief_rejects_multiple_speakers():
             language="en",
             style=PodcastStyle.MONOLOGUE,
             speakers=[_speaker(0), _speaker(1, voice_id="kokoro:af_bella")],
-            duration=DurationTarget(min_minutes=5, max_minutes=10),
+            duration=DurationTarget(min_seconds=300, max_seconds=600),
         )
 
 
 def test_duration_rejects_an_inverted_range():
     """A max below the min is a user error caught at the brief gate."""
     with pytest.raises(ValidationError):
-        DurationTarget(min_minutes=20, max_minutes=10)
+        DurationTarget(min_seconds=1200, max_seconds=600)
 
 
 def test_duration_midpoint_is_where_drafting_aims():
-    assert DurationTarget(min_minutes=10, max_minutes=20).midpoint_minutes == 15
+    assert DurationTarget(min_seconds=600, max_seconds=1200).midpoint_seconds == 900
+    assert DurationTarget(min_seconds=600, max_seconds=1200).midpoint_minutes == 15
+
+
+def test_duration_loads_legacy_minute_fields_from_json():
+    duration = DurationTarget.model_validate({"min_minutes": 10, "max_minutes": 20})
+    assert duration.min_seconds == 600
+    assert duration.max_seconds == 1200
 
 
 def test_blank_focus_becomes_absent():
@@ -117,7 +124,7 @@ def test_blank_focus_becomes_absent():
     spec = PodcastSpec(
         language="en",
         speakers=[_speaker(0)],
-        duration=DurationTarget(min_minutes=5, max_minutes=10),
+        duration=DurationTarget(min_seconds=300, max_seconds=600),
         focus="   ",
     )
     assert spec.focus is None
@@ -127,7 +134,7 @@ def test_speaker_for_returns_the_speaker_bound_to_a_slot():
     spec = PodcastSpec(
         language="en",
         speakers=[_speaker(0), _speaker(1, voice_id="kokoro:af_bella")],
-        duration=DurationTarget(min_minutes=5, max_minutes=10),
+        duration=DurationTarget(min_seconds=300, max_seconds=600),
     )
     assert spec.speaker_for(1).voice_id == "kokoro:af_bella"
 
@@ -136,7 +143,7 @@ def test_speaker_for_raises_when_no_speaker_matches():
     spec = PodcastSpec(
         language="en",
         speakers=[_speaker(0)],
-        duration=DurationTarget(min_minutes=5, max_minutes=10),
+        duration=DurationTarget(min_seconds=300, max_seconds=600),
     )
     with pytest.raises(KeyError):
         spec.speaker_for(99)
diff --git a/surfsense_backend/tests/unit/services/test_quota_checked_vision_llm.py b/surfsense_backend/tests/unit/services/test_quota_checked_vision_llm.py
index 9e35b6f9c..0f5dd531f 100644
--- a/surfsense_backend/tests/unit/services/test_quota_checked_vision_llm.py
+++ b/surfsense_backend/tests/unit/services/test_quota_checked_vision_llm.py
@@ -105,8 +105,7 @@ async def test_ainvoke_propagates_quota_insufficient_error(monkeypatch):
     async def _denying_billable_call(**_kwargs):
         raise QuotaInsufficientError(
             usage_type="vision_extraction",
-            used_micros=5_000_000,
-            limit_micros=5_000_000,
+            balance_micros=5_000_000,
             remaining_micros=0,
         )
         yield  # unreachable but required for asynccontextmanager type
diff --git a/surfsense_backend/tests/unit/tasks/test_video_presentation_billing.py b/surfsense_backend/tests/unit/tasks/test_video_presentation_billing.py
index 423b64ddb..7183024ed 100644
--- a/surfsense_backend/tests/unit/tasks/test_video_presentation_billing.py
+++ b/surfsense_backend/tests/unit/tasks/test_video_presentation_billing.py
@@ -98,8 +98,7 @@ async def _denying_billable_call(**kwargs):
     _CALL_LOG.append(kwargs)
     raise QuotaInsufficientError(
         usage_type=kwargs.get("usage_type", "?"),
-        used_micros=5_000_000,
-        limit_micros=5_000_000,
+        balance_micros=5_000_000,
         remaining_micros=0,
     )
     yield SimpleNamespace()  # pragma: no cover
diff --git a/surfsense_web/components/tool-ui/podcast/brief-review.tsx b/surfsense_web/components/tool-ui/podcast/brief-review.tsx
index 3473b64d6..d3962374a 100644
--- a/surfsense_web/components/tool-ui/podcast/brief-review.tsx
+++ b/surfsense_web/components/tool-ui/podcast/brief-review.tsx
@@ -15,7 +15,9 @@ import {
 } from "@/components/ui/select";
 import { Textarea } from "@/components/ui/textarea";
 import {
+	MAX_DURATION_SECONDS,
 	MAX_SPEAKERS,
+	MIN_DURATION_SECONDS,
 	type PodcastSpec,
 	type PodcastStyle,
 	podcastStyle,
@@ -55,6 +57,9 @@ interface BriefReviewProps {
  */
 export function BriefReview({ podcast, spec }: BriefReviewProps) {
 	const [draft, setDraft] = useState<PodcastSpec>(spec);
+	const [durationUnit, setDurationUnit] = useState<DurationUnit>(() =>
+		defaultDurationUnit(spec.duration.max_seconds),
+	);
 	const [voices, setVoices] = useState<VoiceOption[] | null>(null);
 	const [isSubmitting, setIsSubmitting] = useState(false);
 
@@ -63,6 +68,7 @@ export function BriefReview({ podcast, spec }: BriefReviewProps) {
 	// biome-ignore lint/correctness/useExhaustiveDependencies: reset only when the server version moves
 	useEffect(() => {
 		setDraft(spec);
+		setDurationUnit(defaultDurationUnit(spec.duration.max_seconds));
 	}, [podcast.specVersion]);
 
 	useEffect(() => {
@@ -304,39 +310,72 @@ export function BriefReview({ podcast, spec }: BriefReviewProps) {
 				))}
 			</div>
 
-			<div className="grid grid-cols-2 gap-4">
-				<div className="flex flex-col gap-2">
-					<Label htmlFor="podcast-min-minutes">Min length (minutes)</Label>
-					<Input
-						id="podcast-min-minutes"
-						type="number"
-						min={1}
-						value={draft.duration.min_minutes}
-						onChange={(e) =>
-							setDraft((current) => ({
-								...current,
-								duration: { ...current.duration, min_minutes: Number(e.target.value) || 1 },
-							}))
-						}
-					/>
+			<div className="flex flex-col gap-2">
+				<div className="flex items-center justify-between gap-3">
+					<Label>Target length</Label>
+					<Select
+						value={durationUnit}
+						onValueChange={(value) => setDurationUnit(value as DurationUnit)}
+					>
+						<SelectTrigger className="w-[7.5rem]" aria-label="Length unit">
+							<SelectValue />
+						</SelectTrigger>
+						<SelectContent>
+							<SelectItem value="seconds">Seconds</SelectItem>
+							<SelectItem value="minutes">Minutes</SelectItem>
+							<SelectItem value="hours">Hours</SelectItem>
+						</SelectContent>
+					</Select>
 				</div>
-				<div className="flex flex-col gap-2">
-					<Label htmlFor="podcast-max-minutes">Max length (minutes)</Label>
-					<Input
-						id="podcast-max-minutes"
-						type="number"
-						min={draft.duration.min_minutes}
-						value={draft.duration.max_minutes}
-						onChange={(e) =>
-							setDraft((current) => ({
-								...current,
-								duration: {
-									...current.duration,
-									max_minutes: Number(e.target.value) || current.duration.min_minutes,
-								},
-							}))
-						}
-					/>
+				<div className="grid grid-cols-2 gap-4">
+					<div className="flex flex-col gap-2">
+						<Label htmlFor="podcast-min-length">Min</Label>
+						<Input
+							id="podcast-min-length"
+							type="number"
+							min={durationUnitBounds(durationUnit).min}
+							max={durationUnitBounds(durationUnit).max}
+							step={durationInputStep(durationUnit)}
+							value={formatDurationForUnit(draft.duration.min_seconds, durationUnit)}
+							onChange={(e) => {
+								const seconds = clampDurationSeconds(
+									fromUnitValue(Number(e.target.value), durationUnit),
+								);
+								setDraft((current) => ({
+									...current,
+									duration: { ...current.duration, min_seconds: seconds },
+								}));
+							}}
+						/>
+					</div>
+					<div className="flex flex-col gap-2">
+						<Label htmlFor="podcast-max-length">Max</Label>
+						<Input
+							id="podcast-max-length"
+							type="number"
+							min={secondsToUnitValue(draft.duration.min_seconds, durationUnit)}
+							max={durationUnitBounds(durationUnit).max}
+							step={durationInputStep(durationUnit)}
+							value={formatDurationForUnit(draft.duration.max_seconds, durationUnit)}
+							onChange={(e) => {
+								const parsed = Number(e.target.value);
+								const fallback = secondsToUnitValue(
+									draft.duration.min_seconds,
+									durationUnit,
+								);
+								const seconds = clampDurationSeconds(
+									fromUnitValue(
+										Number.isFinite(parsed) ? parsed : fallback,
+										durationUnit,
+									),
+								);
+								setDraft((current) => ({
+									...current,
+									duration: { ...current.duration, max_seconds: seconds },
+								}));
+							}}
+						/>
+					</div>
 				</div>
 			</div>
 
@@ -365,7 +404,9 @@ export function BriefReview({ podcast, spec }: BriefReviewProps) {
 				<Button
 					type="button"
 					onClick={handleApprove}
-					disabled={isSubmitting || draft.duration.max_minutes < draft.duration.min_minutes}
+					disabled={
+						isSubmitting || draft.duration.max_seconds < draft.duration.min_seconds
+					}
 				>
 					{isSubmitting ? <Loader2 className="size-4 animate-spin" /> : null}
 					{isDirty ? "Approve changes & draft transcript" : "Approve & draft transcript"}
@@ -377,6 +418,50 @@ export function BriefReview({ podcast, spec }: BriefReviewProps) {
 
 /** The current selection stays listed even when it no longer matches the
  * language filter, so the Select never renders an orphaned value. */
+type DurationUnit = "seconds" | "minutes" | "hours";
+
+function defaultDurationUnit(maxSeconds: number): DurationUnit {
+	if (maxSeconds >= 3600) return "hours";
+	if (maxSeconds >= 60) return "minutes";
+	return "seconds";
+}
+
+function secondsToUnitValue(seconds: number, unit: DurationUnit): number {
+	if (unit === "minutes") return seconds / 60;
+	if (unit === "hours") return seconds / 3600;
+	return seconds;
+}
+
+function fromUnitValue(value: number, unit: DurationUnit): number {
+	if (!Number.isFinite(value)) return MIN_DURATION_SECONDS;
+	if (unit === "minutes") return value * 60;
+	if (unit === "hours") return value * 3600;
+	return value;
+}
+
+function formatDurationForUnit(seconds: number, unit: DurationUnit): number {
+	const raw = secondsToUnitValue(seconds, unit);
+	if (unit === "seconds") return Math.round(raw);
+	return Math.round(raw * 100) / 100;
+}
+
+function durationInputStep(unit: DurationUnit): number {
+	if (unit === "hours") return 0.1;
+	return 1;
+}
+
+function durationUnitBounds(unit: DurationUnit): { min: number; max: number } {
+	return {
+		min: formatDurationForUnit(MIN_DURATION_SECONDS, unit),
+		max: formatDurationForUnit(MAX_DURATION_SECONDS, unit),
+	};
+}
+
+function clampDurationSeconds(value: number): number {
+	if (!Number.isFinite(value)) return MIN_DURATION_SECONDS;
+	return Math.min(MAX_DURATION_SECONDS, Math.max(MIN_DURATION_SECONDS, Math.round(value)));
+}
+
 function voiceItems(candidates: VoiceOption[], selectedId: string): VoiceOption[] {
 	if (candidates.some((voice) => voice.voice_id === selectedId)) return candidates;
 	return [
diff --git a/surfsense_web/contracts/types/podcast.types.ts b/surfsense_web/contracts/types/podcast.types.ts
index e6332d5b2..6ceb18dfc 100644
--- a/surfsense_web/contracts/types/podcast.types.ts
+++ b/surfsense_web/contracts/types/podcast.types.ts
@@ -47,6 +47,11 @@ export type PodcastStyle = z.infer<typeof podcastStyle>;
 
 export const MAX_SPEAKERS = 6;
 
+export const MAX_DURATION_SECONDS = 24 * 60 * 60;
+export const MIN_DURATION_SECONDS = 15;
+export const DEFAULT_MIN_SECONDS = 20;
+export const DEFAULT_MAX_SECONDS = 30;
+
 export const speakerSpec = z.object({
 	slot: z.number().int().min(0),
 	name: z.string().min(1).max(120),
@@ -55,10 +60,40 @@ export const speakerSpec = z.object({
 });
 export type SpeakerSpec = z.infer<typeof speakerSpec>;
 
-export const durationTarget = z.object({
-	min_minutes: z.number().int().min(1),
-	max_minutes: z.number().int().min(1),
-});
+export const durationTarget = z.preprocess(
+	(raw) => {
+		if (
+			raw &&
+			typeof raw === "object" &&
+			"min_minutes" in raw &&
+			!("min_seconds" in raw)
+		) {
+			const legacy = raw as { min_minutes: number; max_minutes: number };
+			return {
+				min_seconds: legacy.min_minutes * 60,
+				max_seconds: legacy.max_minutes * 60,
+			};
+		}
+		return raw;
+	},
+	z
+		.object({
+			min_seconds: z
+				.number()
+				.int()
+				.min(MIN_DURATION_SECONDS)
+				.max(MAX_DURATION_SECONDS),
+			max_seconds: z
+				.number()
+				.int()
+				.min(MIN_DURATION_SECONDS)
+				.max(MAX_DURATION_SECONDS),
+		})
+		.refine((duration) => duration.max_seconds >= duration.min_seconds, {
+			message: "Max length must be at least min length",
+			path: ["max_seconds"],
+		}),
+);
 export type DurationTarget = z.infer<typeof durationTarget>;
 
 export const podcastSpec = z